JP3589966B2 - Image processing apparatus and image processing method - Google Patents

Description

領域分離機能部３８は、さらに、上記第１画像ブロック、すなわち１３×７画素のマトリクス内の画素に関して、隣接画素間の画像データの差を求め、上記マトリクス内のすべての画素についての画像データの差の総和を算出する（ステップＳ２）。
【００３３】
具体的には、領域分離機能部３８は、下記（２）式に従って読取部１における主走査方向に関する画素間の画像データの差の総和ＴＯＴＡＬ＿ＤＨを求める。また、下記（３）式に従って読取部１における副走査方向に隣接する画素間の画像データの差の総和ＴＯＴＡＬ＿ＤＶを求める。そして、これらを加算することによって、１３×７画素のマトリクス内の画像データの差の総和ＴＯＴＡＬ＿ＳＵＭを算出する（下記（４）式参照）。むろん、ＴＯＴＡＬ＿ＤＨ，ＴＯＴＡＬ＿ＤＶを求めることなく、画素データの差の総和ＴＯＴＡＬ＿ＳＵＭを直接的に求めてもかまわない。
【００３４】

領域分離機能部３８は、また、読取部１における主走査方向に沿った横線が高密度で平行に形成された横細線領域、および読取部１における副走査方向に沿った縦線が平行に高密度で形成された縦細線領域を写真領域から区別するために、注目画素が細線領域（横細線領域または縦細線領域）に属するか否かを判定する。
【００３５】
この判定のために、領域分離機能部３８は、注目画素を含む一定の大きさの第２画像ブロック中の各画素につき、第１方向（たとえば主走査方向）に隣接する画素間の画像データの差を求め、この画像データの差を上記第２画像ブロック内の全画素について累計することにより第１方向濃度差総和を求める。また、領域分離機能部３８は、上記第２画像ブロック中の各画素について、第１方向と直交する第２方向（たとえば副走査方向）に隣接する画素間の画像データの差を求め、この画像データの差を上記第２ブロック内の全画素について累計することにより、第２方向濃度差総和を求める。
【００３６】
具体的には、図５に示されているように、適当なバッファメモリを用いて、注目画素Ｄ７を中心とした画素Ｂ５〜Ｂ９，Ｃ５〜Ｃ９，Ｄ５〜Ｄ９，Ｅ５〜Ｅ９，Ｆ５〜Ｆ９からなる５画素×５ラインのマトリクスが第２画像ブロックとして形成される。そして、下記（５）式に基づいて、第１方向濃度差総和としての主走査方向濃度差総和Ｈ＿ＳＵＭを求め、下記（６）式に基づいて、第２方向濃度差総和としての副走査方向濃度差総和Ｖ＿ＳＵＭを算出する（図３のステップＳ３）。
【００３７】

次に、領域分離機能部３８は、判定閾値テーブル４０（図２参照）を参照することによって、上記ステップＳ１で求められた平均値ＡＶの関数として定められた判定閾値ＴＨ（ＡＶ）（写真領域判定閾値）を読み出す。この読み出された判定閾値ＴＨ（ＡＶ）と、上記１３×７画素のマトリクス内の画像データの差の総和ＴＯＴＡＬ＿ＳＵＭとが比較される（ステップＳ４）。すなわち、画像データの差の総和ＴＯＴＡＬ＿ＳＵＭが、判定閾値ＴＨ（ＡＶ）以上であれば（ステップＳ４のＮＯ）、注目画素が写真領域以外の種類の画像を構成する画素であるものと判定される（ステップＳ７）。したがって、領域分離機能部３８は、これに対応した領域分離信号（たとえば１ビットの信号であって、その一方の値「０」に設定された信号）を出力する。
【００３８】
これに対して、画像データの差の総和TOTAL_SUMが判定閾値ＴＨ(AV)未満であれば（ステップＳ４のＹＥＳ）、注目画素Ｄ７が写真領域に属するとの一応の判定が行われることになる。ただし、この判定が正しいかどうかは、後に説明するステップＳ５，Ｓ６の処理によって、再判定される。
図６は、平均値ＡＶに対応した閾値ＴＨ(AV)の設定例を説明するためのグラフである。この図６には、平均値ＡＶと画像データの差の総和TOTAL_SUMとの関係が示されている。平均値ＡＶに対する閾値ＴＨ(AV)の設定例は、実線の曲線で表されている。また、写真領域における平均値ＡＶと画像データの差の総和TOTAL_SUMとの関係の分布がシンボル「○」で表されていて、さらに、写真領域以外（文字領域、線画領域、および低線数の網点領域）における平均値ＡＶと画像データの差の総和TOTAL_SUMとの関係の分布がシンボル「×」によって表されている。
【００３９】
図６から理解されるように、閾値ＴＨ（ＡＶ）は、シンボル「○」の分布域と、シンボル「×」の分布域との間に位置するように設定されている。したがって、画像データの差の総和ＴＯＴＡＬ＿ＳＵＭが閾値ＴＨ（ＡＶ）未満であることにより、注目画素が写真領域に属するものと判定することができ、画像データの差の総和ＴＯＴＡＬ＿ＳＵＭが平均値ＡＶに応じて設定された閾値ＴＨ（ＡＶ）以上の場合には、注目画素が写真領域以外の領域に属するものと判断できる。
【００４０】
再び図３を参照すると、領域分離機能部３８は、さらに上記第２画像ブロックを構成する５×５画素のマトリクスについて求められた主走査方向濃度差総和Ｈ＿ＳＵＭおよび副走査方向濃度差総和Ｖ＿ＳＵＭの大小関係をステップＳ５，Ｓ６の処理によって調べ、その結果に応じて、注目画素Ｄ７が細線領域に属するか否かを判定する。細線領域に属すると判定された場合には、ステップＳ７の処理によって、注目画素Ｄ７が写真領域に属しないと判定され、それに応じた領域分離信号「０」が生成される。
【００４１】
具体的に説明すると、図７に示されているように、水平方向濃度差総和Ｈ＿ＳＵＭに関して２つの閾値ＨＬ，ＨＨ（ＨＨ＞ＨＬ）が設定され、同様に副走査方向濃度差総和Ｖ＿ＳＵＭに関して、２つの閾値ＶＬ，ＶＨ（ＶＨ＞ＶＬ）が設定されている。たとえば、閾値ＨＬ，ＨＨ；ＶＬ，ＶＨは、次のような値に設定される。
ＨＬ＝５０ ＨＨ＝１５０
ＶＬ＝５０ ＶＨ＝１００
図７に示されているように、副走査方向濃度差総和Ｖ＿ＳＵＭを横軸にとり、主走査方向濃度差総和Ｈ＿ＳＵＭを縦軸にとった平面において、横方向（主走査方向）に沿う細線が密集した横細線領域を構成する画素については、副走査方向濃度差総和Ｖ＿ＳＵＭが大きく、かつ主走査方向濃度差総和Ｈ＿ＳＵＭが小さな領域ＨＦに多くの画素が分布する。また、縦方向（副走査方向）に沿う細線が密集した縦細線領域に属する画素については、副走査方向濃度差総和Ｖ＿ＳＵＭが小さく、かつ主走査方向濃度差総和Ｈ＿ＳＵＭが大きな領域ＶＦに多くの画素が分布している。そして、写真領域とみなすべき網点画像および本来の写真画像については、Ｖ＿ＳＵＭ−Ｈ＿ＳＵＭ平面の中央付近の領域ＰＳに画素の分布が得られる。
【００４２】
そこで、この実施形態では、領域分離機能部３８は、注目画素についての主走査方向濃度差総和Ｈ＿ＳＵＭが閾値ＨＨ以上であって、かつ副走査方向濃度差総和をＶ＿ＳＵＭが閾値ＶＬ以下である場合に（図３のステップＳ５のＹＥＳ）、注目画素が細線領域（この場合は縦細線領域）に属するものと判定する。すなわち、注目画素は写真領域には属さないものと判定されて、このことを表す領域分離信号「０」が生成される（ステップＳ７）。
【００４３】
また、領域分離機能部３８は、注目画素に関する主走査方向濃度差総和Ｈ＿ＳＵＭが閾値ＨＬ以下であって、かつ副走査方向濃度差総和Ｖ＿ＳＵＭが閾値ＶＨ以上である場合に（ステップＳ６のＹＥＳ）、注目画素が細線領域（この場合には横細線領域）に属するものと判定する。すなわち、注目画素が写真領域に属しないと判定され、それに応じた領域分離信号「０」が生成される（ステップＳ７）。
ステップＳ５，Ｓ６の判断がいずれも否定される場合、すなわち、主走査方向濃度差総和Ｈ＿ＳＵＭが閾値ＨＨ以上であるが、副走査方向濃度差総和Ｖ＿ＳＵＭが閾値ＶＬを超えている場合、主走査方向濃度差総和Ｈ＿ＳＵＭが閾値ＨＬ以下であるが、副走査方向濃度差総和Ｖ＿ＳＵＭが閾値ＶＨ未満である場合、および主走査方向濃度差総和Ｈ＿ＳＵＭがＨＬ〜ＨＨの範囲にある場合には、注目画素Ｄ７は細線領域に属さないものと判定される。すなわち、注目画素Ｄ７に対して、写真領域に属するものとの再判定がなされる（ステップＳ８）。この場合、領域分離機能部３８は、たとえば１ビットの領域分離信号を写真領域に対応した値（この実施形態では「１」）に設定する。
【００４４】
このようにして、１３×７画素のマトリクス内の画素間の濃度差の総和に基づく判定（ステップＳ４のＹＥＳ）が、再確認されることになる。
図８は、分離前フィルタ処理部３９の処理を説明するための図である。分離前フィルタ処理部３９は、基本的には、積分フィルタとしての処理を行う積分処理手段を構成している。すなわち、図８（ａ）に示されているように、注目画素ｃ３を中心とした合計１３画素（ａ３，ｂ２〜ｂ４，ｃ１〜ｃ５，ｄ２〜ｄ４，ｅ３）からなる菱形（主走査方向に対して４５度傾斜した正方形）のマトリクス内の画素の画像データを用いて注目画素ｃ３の画像データを平滑化する。
【００４５】
図８(a)のマトリクス内の各画素の画像データに乗じられる係数ｓ１〜ｓ６は、図８(b)に示すように、たとえば、注目画素ｃ３を中心として、主走査方向および副走査方向に関してほぼ対称となるように設定される。ｓ１〜ｓ６の値は、次式のように等しく設定されてもよいし、異なる値に設定されてもよい。すなわち、たとえば、注目画素からの距離に応じて重み付けを異ならせるようにｓ１〜ｓ６の値を定めてもよい。
【００４６】
ｓ１＝ｓ２＝ｓ３＝ｓ４＝ｓ５＝ｓ６＝１
ｓ１〜ｓ６が上式の値に定められている場合には、注目画素ｃ３の画像データは、次の値Ｆ（ｃ３）に変換されることになる。

この実施形態においては、たとえば、中間調処理部３７において写真領域に属する画素に施される改良型誤差拡散処理では、スクリーン角４５度、線数１４１線の網点が形成される。一方、読取部１における読取解像度は、たとえば、６００ｄｐｉ（ｄｏｔ ｐｅｒ ｉｎｃｈ）である。そうすると、１４１線の網点においては、図９に示すように、主走査方向および副走査方向にそれぞれ３画素ずつ離れた斜め４５度のライン上の画素ｐ１，ｐ２に相当する位置に網点が存在することになる。
【００４７】
一般的に、網点化処理する対象は中間処理の際の網点化の線数に近いほどドットが良好に安定化される。これは、画像出力装置において網点化処理する線数はその装置で画像再現能を維持できる範囲で高い線数を選んでいるので、網点化する線数がそれ以上の線数であれば、網点化しても再現するドットの安定性が悪くなるからである。すなわち、画像出力装置の中間調処理における網点化の線数以上の網点を忠実に再現しても安定したドットは形成できない。画像出力装置や中間調処理の線数にもよるが、最近では６００ｄｐｉで処理するのが普通であり、一般的に中間調処理の線数と原稿の細線の線数との差が±５０線以内であればドットの安定性は維持される。
【００４８】
一方、モアレ縞防止の観点からは、中間調処理における網点化の線数と画像処理される原稿の網点線数は近くない方がよい。これは、お互いの線数が近ければ近いほど人間の目につくほどの大きなモアレ縞が発生するからである。一般的に中間調処理の線数の±５０線も離れればモアレ縞はほとんど発生しない。
また、網点化処理されるのは写真などの高線数側のものが普通であるから、これらのことにより中間調処理の線数（網点化処理の線数）とこの線数＋５０線との間に中間調処理における領域分離の境界線数（たとえば、１７５線）を設定して、その線数よりも高い線数領域を写真領域と判断し、それ以下を写真領域以外として除外する。
【００４９】
原稿の網点線数と中間調処理部３７による改良型誤差拡散処理部により形成される網点の線数とが近似していて、線数の差が±５０線の範囲内の小さな値であるとき、いわゆるモアレ縞が出力画像中に形成されるおそれがある。そこで、このような線数範囲の原稿画素については、改良型誤差拡散処理ではなく、量子化閾値を一定値に固定した通常型誤差拡散処理を行うことが適切である。その一方で、改良型誤差拡散処理により形成される網点線数である１４１線と１４１＋５０線との間で定めた境界線数以上の高線数の網点画像については、写真画像とみなして改良型誤差拡散処理を適用することによって、良好な階調再現を実現できる。
【００５０】
よって、改良型誤差拡散処理を行う所定線数（この実施形態では１４１線）とこれよりも＋５０線多い線数（この実施形態では１９１線）との間に境界線数を設定し、その境界線数以上の線数の画像を写真領域、それ以外の画像をその他の領域と領域分離判定する。この境界線数は、分離前フィルタ処理部３９の構成によって決定される。具体的には、図８（ａ）のマトリクスの範囲を大きくするほど干渉の生じる線数が低下して境界線数が低くなる。また、マトリクスの係数ｓ１〜ｓ６に差をつける（たとえば、中心が高くなるように差をつける）と、境界線数が大きくなる。図８（ａ）のマトリクスを用いて係数ｓ１〜ｓ６を等しく定めているこの実施形態の場合には、境界線数は約１７５線となる。
【００５１】
分離前フィルタ処理部３９における上述の積分フィルタは、図９に示されているように、１４１線の網点原稿に関しては、積分フィルタ処理後の画像データにおいて、網点相互間に干渉が生じないように定められている（図９において、平滑化処理後の網点を斜線で示す）。そして、この積分フィルタは、入力画像が１７５線以上の高線数の網点画像の場合に、隣接する網点間に良好な干渉を生じさせる。その結果、領域分離機能部３８では、１７５線以上の高線数の網点画像については、上述の１３×７画素のマトリクス内における画素間の画像データの差の総和ＴＯＴＡＬ＿ＳＵＭが比較的小さいものと判定されて、このような高線数の網点画素を構成する画素が写真領域に属するものと判定されることになる。
【００５２】
念のため付言すると、上述の第１画像ブロックは１３×７画素の大きさのマトリクスである必要はなく、同様に上述の第２画像ブロックは５×５画素の大きさのマトリクスである必要はない。
すなわち、第１画像ブロックの大きさは、当該第１画像ブロック内における上述の平均値ＡＶと、画像データの差の総和ＴＯＴＡＬ＿ＳＵＭとの関係の分布が、上述の図６に示されているように、分離された領域に現れるように設定すればよい。
【００５３】
上述の実施形態において１３×７画素のマトリクスを採用しているのは、６５線の網点画像を構成する網点をブロック内に必ず含む領域を参照するためである。６５線の網点画像は、たとえば新聞の写真印刷に用いられる画像であって、一般的な印刷物において用いられる最小の網点線数である。したがって、６５線以上の網点領域を網点領域として（写真領域以外として）分離することができれば、実用上充分である。
【００５４】
この目的のためには、６５線の網点画像を構成する網点が参照範囲（第１画像ブロック）内に最低１つ存在しなければならない。この場合、図１０に示されているように、網点Ｄが４５度のスクリーン角で形成されるとすると、主走査方向に沿って隣接する網点間の間隔は、√２／６５インチであり、副走査方向に沿う網点Ｄ間の間隔は１／（６５・√２）インチである。
読取部１における読取解像度が６００ｄｐｉであるとすると、主走査方向に沿う隣接網点Ｄ間の間隔は、１３（≒６００×√２／６５）画素となり、副走査方向に沿った網点Ｄ間の間隔が７（≒６．５３＝６００／（６５・√２））画素となる。したがって、上述の実施形態で説明したように、第１画像ブロックの大きさを１３×７画素のマトリクスに設定することによって、６５線以上の線数の網点領域を写真領域以外の領域として分離することができる。
【００５５】
このように、第１画像ブロックの大きさは、分離対象の網点線数および読取部１の読取解像度に依存しており、１３×７画素のマトリクスが単なる一例であることが理解される。
第１画像ブロックを形成するためのバッファメモリの容量を少なくするためには、対象とする最低線数の網点画像の網点が参照マトリクス内に１つ存在するように、第１画像ブロックの大きさを定めることが好ましい。しかし、第１画像ブロック中に１つよりも多く（整数である必要はない）の網点が存在するように第１画像ブロックの大きさを設定することもできる。すなわち、６５線以上の網点領域を写真領域以外であると判定することを目的とする場合に、１３×７画素のマトリクスよりも大きなマトリクスを第１画像ブロックとして用いてもよい。
【００５６】
上述の第２画像ブロックについても同様であって、主走査方向濃度差総和Ｈ＿ＳＵＭと副走査方向濃度差総和Ｖ＿ＳＵＭとの関係の分布が、上述の図７に示されているように、写真とみなすべき画像と、細線領域画像とで分離されるように画素マトリクスの大きさを定めればよい。したがって、たとえば、第２画像ブロックは、５×５画素よりも小さなまたは大きな正方形の画素マトリクスであってもよいし、主走査方向と副走査方向とで画素数の異なる矩形またはその他の形状のマトリクスであってもよい。たとえば、第１画像ブロックを構成する１３×７画素のマトリクスをそのまま第２画像ブロックとして採用し、ＴＯＴＡＬ＿ＤＨおよびＴＯＴＡＬ＿ＤＶをそれぞれＨ＿ＳＵＭおよびＶ＿ＳＵＭに代えて用いることもできる。
【００５７】
ただし、参照マトリクスを大きくすると、各ラインにおいて初期に処理される画素については、参照可能な画像データが少ないので、領域分離判定に誤りが生じるおそれがある。この観点からは、第１画像ブロックおよび第２画像ブロックともに、必要最小限の大きさのマトリクスを採用しておくことが好ましい。
なお、主走査方向両端部および副走査方向両端部では、第１画像ブロックまたは第２画像ブロックを構成する画素が不足することになる。たとえば、各ラインにおいて最初の画素を注目画素として処理するときには、主走査方向上流側には画素が存在していない。このようなときには、１３×７画素のマトリクスまたは５×５画素のマトリクス内において、存在しないマトリクス位置の画素の画像データを「０」と見なして、上述の領域分離判定が行われる。
【００５８】
図１１は、中間調処理部３７の機能的な構成を説明するためのブロック図である。中間調処理部３７は、領域分離機能部３８から与えられる領域分離信号と、注目画素の座標位置を表す注目画素座標と、注目画素の画像データ（この実施形態では、出力γ補正処理後の画像データ）とに基づき、誤差拡散処理用のパラメータを生成するパラメータ生成処理部５１を備えている。また、中間調処理部３７は、出力γ補正処理部３６から与えられる画像データと、パラメータ生成処理部５１から与えられる各種のパラメータとに基づき、入力画像データに対して誤差拡散処理演算を行う誤差拡散処理部５２を備えている。
【００５９】
パラメータ生成処理部５１は、後述する構成により、入力画像データを量子化（この実施形態では二値化または四値化）する際の閾値Ｖ１，Ｖ２，Ｖ３またはＦ１，Ｆ２，Ｆ３と、画像データを量子化する際に生じる発生誤差（量子化誤差）を補正するための発生誤差補正値ＥＣとを生成して、誤差拡散処理部５２に与えるようになっている。
誤差拡散処理部５２は、入力画像データに対して、累積誤差メモリ６２に記憶されている累積誤差を加算する誤差加算処理部６１を備えている。累積誤差メモリ６２には、量子化処理が既に終了した周辺の画素から注目画素に分配される誤差が累積されて記憶されている。たとえば、図１２（ａ）に示されているように、注目画素Ｘに対しては、その周辺の既に量子化された画素ａ，ｂ，ｃ，ｄ，ｅ，ｆから予め定める誤差拡散係数（たとえば、１／４または１／８）を乗じた発生誤差が累積されて加算されることになる。この場合の累積誤差は、たとえば、次式により表される。
【００６０】

この累積誤差が、それぞれの画素に対応づけて累積誤差メモリ６２に記憶されており、誤差拡散処理対象の注目画素に対応した累積誤差が読み出されて、誤差加算処理部６１により、画像データに加算されることになる。
誤差拡散処理部５２は、さらに、誤差加算処理後の画像データを、パラメータ生成処理部５１から与えられる、それぞれ通常の誤差拡散処理と改良型誤差拡散処理とに対応した３つの量子化閾値Ｖ１，Ｖ２，Ｖ３またはＦ１，Ｆ２，Ｆ３に基づいて量子化する量子化処理部６３を備えている。この量子化処理部６３は、この実施形態では、画像データを、０，１，２，３の四値を表す２ビットのデータに変換する。
【００６１】
量子化閾値Ｖ１，Ｖ２，Ｖ３は、画像平面に沿う二次元方向に関して周期変動するように設定される変動閾値であり、それらの間には、Ｖ１≦Ｖ２≦Ｖ３なる大小関係が成立する。また、量子化閾値Ｆ１，Ｆ２，Ｆ３は、固定閾値であってＦ１≦Ｆ２≦Ｆ３（この実施形態では、とくにＦ１＝Ｆ２＝Ｆ３）なる大小関係が成立する。パラメータ生成処理部５１は、変動閾値Ｖ１，Ｖ２，Ｖ３または固定閾値Ｆ１，Ｆ２，Ｆ３のいずれかを生成する。これを受けて量子化処理部６３は、３つの量子化閾値Ｖ１，Ｖ２，Ｖ３またはＦ１，Ｆ２，Ｆ３に基づいて、次の場合分け判定Ｉ，ＩＩ，ＩＩＩ，ＩＶをこの順に実行することにより、画像データを０〜３に量子化（四値化）する。場合分け判定をＩ〜ＩＶの順に実行する過程で、いずれかの条件に合致すれば、以後の判定は行わず、量子化された値（出力データ）を確定する。
【００６２】

【００６３】
IV．Ｖ１またはＦ１≧誤差加算処理後の画像データのとき、
注目画素を「０」に量子化する。
・・・・・・（９）
このようにして、量子化閾値Ｖ１，Ｖ２，Ｖ３またはＦ１，Ｆ２，Ｆ３が異なる値を有する場合には、注目画素の画像データは、「０」〜「３」の四値に量子化される。また、閾値Ｖ１，Ｖ２，Ｖ３またはＦ１，Ｆ２，Ｆ３が等しい値を有する場合には、上述の場合分け判定Ｉ〜IVを順に実行することにより、注目画素の画像データは「０」または「３」に二値化されることになる。
【００６４】
この量子化処理部６３による量子化処理の結果生じた発生誤差が、誤差演算部６４によって求められる。この誤差演算部６４は、入力画像データの濃度階調が０ないし２５５の２５６階調で表されている場合に、出力データ（量子化結果）に応じて、次式に従って発生誤差を演算する。

出力データ「３」は黒画素に対応しており、最大階調値「２５５」に相当する。また、出力データ「０」は白画素に対応しており、最小階調値「０」に相当する。そして、出力データ「１」「２」は、いわばグレーの画素に対応していると言うことができ、それぞれ第１中間階調値「８５」および第２中間階調値「１７０」に相当している。ただし、ここでは、第１中間階調値および第２中間階調値は、０〜２５５の全階調区間を４等分する階調値に選択している。
【００６５】
誤差演算部６４によって演算された発生誤差には、誤差補正部６５によって補正処理が施されることになる。この補正処理は、下記（１１）式に示すように、誤差演算部６４によって求められた発生誤差から、パラメータ生成処理部５１から与えられる発生誤差補正値ＥＣを減算することによって達成される。この補正後の発生誤差に対しては、誤差分配処理部６６による分配処理が行われ、この分配処理の結果としての累積誤差が、累積誤差メモリ６２に格納されることになる。
【００６６】
補正後の発生誤差＝（１０）式の発生誤差−ＥＣ ・・・・・・（１１）
誤差分配処理部６６における処理は、図１２（ａ）に示された誤差の累積と反対の処理となる。すなわち、注目画素Ｘにおいて発生した発生誤差は、図１２（ｂ）に示すように予め定める誤差拡散係数（この例では１／４または１／８）を乗じて、注目画素Ｘの周辺にあって、未だ二値化されていない周辺画素に分配される。
【００６７】
図１３は、パラメータ生成処理部５１の機能的な構成を説明するためのブロック図である。パラメータ生成処理部５１は、領域分離機能部３８から与えられる領域分離信号に基づいて、量子化閾値を一定値に固定した通常型誤差拡散処理に対応したパラメータ、または量子化閾値を画像平面において２次元的に周期変動させて設定する改良型誤差拡散処理のためのパラメータを生成する。すなわち、領域分離信号が、注目画素が写真領域に属することを表す値「１」である場合には、改良型誤差拡散処理のためのパラメータを生成し、領域分離信号が、注目画素が写真領域以外の領域に属することを表す値「０」の場合には通常型誤差拡散処理のためのパラメータを生成する。
【００６８】
このような機能のために、パラメータ生成処理部５１は、通常型誤差拡散処理のための固定閾値Ｆ１，Ｆ２，Ｆ３を生成する固定閾値設定部７１と、改良型誤差拡散処理のための変動閾値Ｖ１，Ｖ２，Ｖ３を設定する変動閾値設定部７２と、領域分離信号に基づいて固定閾値Ｆ１，Ｆ２，Ｆ３または変動閾値Ｖ１，Ｖ２，Ｖ３のいずれかを選択して、誤差拡散処理部５２に与える閾値選択部７３と、発生誤差の補正のための発生誤差補正値ＥＣを演算して、誤差拡散処理部５２に与える補正値演算部７４とを備えている。
【００６９】
補正値演算部７４は、領域分離信号および変動閾値設定部７２から与えられる変動閾値Ｖ１，Ｖ２，Ｖ３に基づいて、発生誤差補正値ＥＣを演算する。具体的には、領域分離信号が「０」であって、注目画素が写真領域以外に属する画素である場合には、発生誤差補正値ＥＣを一定値（この実施形態では「０」）に設定する。その一方で、領域分離信号の値が「１」である場合には、変動閾値設定部７２から与えられる変動閾値Ｖ１，Ｖ２，Ｖ３の値に応じて、発生誤差補正値ＥＣを可変させる。
【００７０】
固定閾値設定部７１が生成する固定閾値Ｆ１，Ｆ２，Ｆ３は、この実施形態では、同一の値（たとえば０〜２５５の中間値である「１２７」）に設定されている。この固定閾値Ｆ１，Ｆ２，Ｆ３が誤差拡散処理部５２に与えられると、この誤差拡散処理部５２の量子化処理部６３は、入力画像データを「０」または「３」のいずれかの値に二値化する。
変動閾値設定部７２は、画像平面において２次元的に周期変動する量子化閾値を設定するための変動閾値マトリクスを記憶した変動閾値マトリクスメモリ８０と、注目画素座標に基づいて、注目画素が変動閾値マトリクス内のいずれの位置に該当するかを見いだし、当該位置における変動閾値マトリクスの値を変動閾値選択情報として生成する変動閾値選択情報生成部８５と、この変動閾値選択情報と注目画素の画像データ値とに基づき、変動閾値Ｖ１，Ｖ２，Ｖ３を生成する変動閾値生成部８６とを備えている。
【００７１】
変動閾値生成部８６は、変動閾値Ｖ１を生成するための変動閾値テーブルＴ１が格納された変動閾値メモリ８１と、変動閾値Ｖ２を生成するための変動閾値テーブルＴ２が格納された変動閾値メモリ８２と、変動閾値Ｖ３を生成するための変動閾値テーブルＴ３が格納された変動閾値メモリ８３とを備えている。さらに、変動閾値生成部８６は、変動閾値選択情報および注目画素の画像データの値に基づいて、変動閾値メモリ８１，８２，８３から、該当する変動閾値Ｖ１，Ｖ２，Ｖ３を選択して閾値選択部７３に与える変動閾値選択部８４を備えている。
【００７２】
閾値選択部７３は、領域分離信号が写真領域に対応する値「１」である場合には変動閾値Ｖ１，Ｖ２，Ｖ３を選択して誤差拡散処理部５２に与える。また、領域分離信号が写真領域以外であること表す値「０」である場合には、固定閾値設定部７１により設定された固定閾値Ｆ１，Ｆ２，Ｆ３を誤差拡散処理部５２に与える。
図１４（ａ）は、変動閾値マトリクスメモリ８０に格納された変動閾値マトリクスの一例を示す図であり、図１４（ｂ）は、この変動閾値マトリクスの各マトリクス位置に対応した変動閾値マトリクスメモリ８０のアドレスを示す図である。図１４（ａ）には、画像データを四値化する場合を考慮した変動閾値マトリクスが示されており、各マトリクス値は、０，１，２，３のうちのいずれかの値に設定されている。閾値マトリクス内において小さなマトリクス値に対応する画素においては、比較的小さな量子化閾値が設定され、大きなマトリクス値に該当する画素に対しては比較的大きな量子化閾値が設定される。したがって、注目画素に該当するマトリクス値が小さければ、この注目画素は高濃度画素（黒画素）に量子化される可能性が高く、注目画素に該当するマトリクス値が大きければ、この注目画素は低濃度画素（白画素）に量子化される可能性が高くなる。
【００７３】
変動閾値Ｖ１，Ｖ２，Ｖ３を用いて画像を量子化する処理は、図１４（ａ）に示す変動閾値マトリクスを画像平面上において２次元方向に敷き詰めた状況を仮想することによって理解される。すなわち、この場合に、量子化処理部６３（図１１参照）における量子化閾値は、主走査方向および副走査方向（すなわち２次元方向）に周期変動することになる。したがって、量子化された後の画像には、周期的にドット（網点）が現れることになり、結果として網点化処理を行うことができる。
【００７４】
また、図１４（ａ）の例では、マトリクス値「０」（量子化閾値の変動周期の鞍点に対応）をマトリクス値「１」で取り囲んだ３×３画素の小マトリクスと、マトリクス値「３」（量子化閾値の変動周期の頂点に対応）をマトリクス値「２」で取り囲んだ３×３画素の小マトリクスとが、主走査方向および副走査方向に交互に配置されている。これにより、スクリーン角４５度の網点が形成されることになる。読取部１の読取解像度が６００ｄｐｉであれば、網点線数は１４１線になる。
【００７５】
図１４（ａ）の変動閾値マトリクスは、６×６画素に設定されているが、５×５画素や７×７画素などの他のマトリクスサイズを採用することもできる。たとえば、ｎ×ｎ画素のマトリクスのマトリクスサイズを「ｎ」と定義すると、注目画素に対応した変動閾値マトリクス内の位置、すなわち変動閾値マトリクスメモリ８０のアドレスは、注目画素の座標（主走査座標および副走査座標）を用いて、次式により得ることができる。
【００７６】
アドレス＝（主走査座標÷ｎの余り＋副走査座標÷ｎの余り）×８ ・・・・・・（１２）
この式は、変動閾値マトリクスの最大マトリクスサイズが「８」である場合を想定したものであって（図１４（ｂ）参照）、変動閾値マトリクスメモリ８０のアドレスの指定は、この変動閾値マトリクスメモリ８０の構成に応じて適宜変更されることになる。
【００７７】
変動閾値選択情報生成部８５は、注目画素の座標から上記（１２）式に基づいて変動閾値マトリクスのマトリクス値を読み出し、これを変動閾値選択情報として変動閾値選択部８４に与える。変動閾値選択部８４は与えられた変動閾値選択情報に基づいて、変動閾値メモリ８１，８２，８３の変動閾値テーブルＴ１，Ｔ２，Ｔ３から、適切な変動閾値Ｖ１，Ｖ２，Ｖ３を読み出す。
図１５は、変動閾値テーブルＴ１，Ｔ２，Ｔ３の設定例を示す図である。変動閾値Ｖ１，Ｖ２，Ｖ３は、Ｖ１≦Ｖ２≦Ｖ３の関係が満たされるように設定されている。
【００７８】
より具体的には、この実施形態においては、変動閾値マトリクスのマトリクス値「０」に対応する変動閾値Ｖ１（０），Ｖ２（０），Ｖ３（０）については、原則として、Ｖ１（０）＝Ｖ２（０）＝Ｖ３（０）となるように変動閾値Ｖ１（０），Ｖ２（０），Ｖ３（０）が定められている。また、変動閾値マトリクスのマトリクス値「２」および「３」についても同様に、Ｖ１（２）＝Ｖ２（２）＝Ｖ３（２）およびＶ１（３）＝Ｖ２（３）＝Ｖ３（３）なる関係が原則として成立するように変動閾値Ｖ１，Ｖ２，Ｖ３が定められている。
【００７９】
そして、変動閾値マトリクス内における特定のマトリクス値であるマトリクス値「１」については、原則として、変動閾値Ｖ１（１），Ｖ２（１），Ｖ３（１）について、Ｖ１（１）＜Ｖ２（１）＜Ｖ３（１）なる関係が成立するように、変動閾値Ｖ１，Ｖ２，Ｖ３が定められている。
さらには、変動閾値Ｖ１，Ｖ２，Ｖ３のいずれについても、原則として、変動閾値マトリクスのマトリクス値が小さいほど、小さな変動閾値が割り当てられている。すなわち、次式の関係が成立するように変動閾値Ｖ１，Ｖ２，Ｖ３が定められている。
【００８０】
Ｖ１（０）＜Ｖ１（１）＜Ｖ１（２）＜Ｖ１（３）
Ｖ２（０）＜Ｖ２（１）＜Ｖ２（２）＜Ｖ２（３）
Ｖ３（０）＜Ｖ３（１）＜Ｖ３（２）＜Ｖ３（３） ・・・・・・（１２）
誤差拡散処理部５２の量子化処理部６３は、変動閾値Ｖ１，Ｖ２，Ｖ３が量子化閾値として与えられた場合に、変動閾値マトリクスのマトリクス値が０，２または３の場合には、変動閾値Ｖ１，Ｖ２，Ｖ３がいずれも等しいので、入力画像データを「０」または「３」のいずれかに二値化する。これに対して、変動閾値マトリクスのマトリクス値が「１」に該当する注目画素に対しては、変動閾値Ｖ１，Ｖ２，Ｖ３が異なる値を有しているので、入力画像データを０ないし３のいずれかの値に四値化する。
【００８１】
図１５において画像データ値の０〜１５の範囲に注目すると理解されるように、この実施形態では、注目画素の画像データの値（濃度値）に応じて、変動閾値Ｖ１，Ｖ２，Ｖ３を可変設定するようになっている。具体的には、画像データの値が０〜７の範囲の場合には、変動閾値Ｖ１，Ｖ２，Ｖ３は、変動閾値マトリクスのマトリクス値に関わらずに、０〜２５５の階調の中間値「１２７」に固定されている。これによって、量子化閾値を周期変動させない通常型誤差拡散処理が行われることになり、誤差拡散処理部５２の量子化処理部６３は、入力画像データを「０」または「３」に二値化する。
【００８２】
また、注目画素の画像データの値が８〜１５の範囲にある場合には、変動閾値Ｖ１，Ｖ２，Ｖ３は、変動閾値マトリクスのマトリクス値に応じた変動幅が画像データの値が１６以上の場合よりも小さくなるように設定される。
このように、極低濃度領域においては、変動閾値マトリクスのマトリクス値に応じた閾値変化を小さくするか、または無くすことによって、ドット集中を緩和または解消することができる。これにより、写真画像中の低濃度領域については、出力画像中において大きなドットが目立つことがなく、美しい低濃度階調画像を再生することができる。それとともに、量子化閾値の変動に起因するテクスチャの発生を抑制できる。
【００８３】
変動閾値選択部８４は、注目画素の画像データと、当該注目画素の変動閾値マトリクス内の位置に該当するマトリクス値（変動閾値選択情報）とに基づいて、変動閾値Ｖ１，Ｖ２，Ｖ３を変動閾値メモリ８１，８２，８３から読み出し、閾値選択部７３に与える。
補正値演算部７４は、領域分離信号の値が「０」であって、固定閾値設定部７１より設定される固定閾値Ｆ１，Ｆ２，Ｆ３が用いられる場合には、発生誤差補正値ＥＣを「０」に設定する。一方、領域分離信号が写真領域であることを表す値「１」である場合には、補正値演算部７４は、閾値選択部７３から変動閾値Ｖ１，Ｖ２，Ｖ３を得て、次式に基づいて、発生誤差補正値ＥＣを演算する。
【００８４】
発生誤差補正値＝（Ｖ１＋Ｖ２＋Ｖ３）÷３−１２７ ・・・・・・（１３）
すなわち、変動閾値Ｖ１，Ｖ２，Ｖ３の平均値の中間階調値「１２７」（全階調範囲を２分する中間階調値）に対する偏差が発生誤差補正値ＥＣとして求められ、誤差補正部６５に与えられる。誤差補正部６５は、誤差演算部６４で演算された発生誤差から発生誤差補正値ＥＣを引くことによって、補正後の発生誤差を演算する（上記(11)式参照）。
【００８５】
このようにして、画像平面において２次元方向に周期変動する変動閾値Ｖ１，Ｖ２，Ｖ３を用いる場合には、周辺の画素に分配される発生誤差が、量子化閾値の変動の影響を受けることがないように補正される。この補正処理は、結局、発生誤差の算出基準値を量子化閾値の変動に応じて変更する処理にほかならない。すなわち、上記（１０）式における最小階調値「０」、第１中間階調値「８５」、第２中間階調値「１７０」および最大階調値「２５５」は、発生誤差の算出基準値である。発生誤差を補正するための上記（１１）式の処理は、結局、発生誤差の算出基準値「０」，「８５」，「１７０」，「２５５」を発生誤差補正値ＥＣで補正し、この補正後の算出基準値「０−ＥＣ」，「８５−ＥＣ」，「１７０−ＥＣ」，「２５５−ＥＣ」を基準として量子化誤差を求める処理に他ならない。発生誤差補正値ＥＣは、上記（１３）式から理解されるとおり、変動閾値Ｖ１，Ｖ２，Ｖ３の周期変動とともに周期変動するから、結局、補正後の算出基準値は、変動閾値Ｖ１，Ｖ２，Ｖ３と同位相で周期変動することになる。これによって、大きな発生誤差が周辺の画素に分配されることを防止できるので、ドットを効率的に集中させて良好な網点を形成することができる。
【００８６】
図１６は、量子化閾値の周期変動とそれに応じた誤差算出基準値の周期変動とを示す図である。量子化閾値（変動閾値）Ｖ１，Ｖ２，Ｖ３が図１６において太線で示すように周期的に変動すると、中間階調値「１２７」に対する量子化閾値の偏差である発生誤差補正値ＥＣは、図１６中の矢印で示すように変化する。この場合、下向きの矢印は発生誤差補正値ＥＣが負の値であることを表し、上向きの矢印は、発生誤差補正値ＥＣが正の値であることを表す。
【００８７】
たとえば、変動閾値マトリクス値「０」に対応した変動閾値Ｖ１＝Ｖ２＝Ｖ３が中間値「１２７」よりも小さく設定されているとする。この場合に、中間値「１２７」よりも小さな値を有する注目画素Ｘは、変動閾値Ｖ１，Ｖ２，Ｖ３よりもその値が大きいときに、黒画素に相当する値「３」に量子化される。このとき、誤差演算部６４は、補正前の発生誤差として、次式に示す値を演算する。
補正前の発生誤差＝注目画素Ｘの累積誤差加算後の値−２５５ ・・・・・・（１４）
したがって、この場合には、「１２７」よりも大きな絶対値の発生誤差が生じることになる。この補正前の発生誤差は、発生誤差補正値ＥＣを減じることによって、絶対値の小さな値に補正される。この補正の結果、補正後の発生誤差は、量子化閾値Ｖ１，Ｖ２，Ｖ３に応じた小さな値に補正される。この補正は、発生誤差の算出基準値を最大階調値「２５５」よりも小さな値Ｒ（３）に補正することと等価である。
【００８８】
なお、図１６において、注目画素の画像データが出力データ「０」，「１」，「２」，「３」に量子化された場合の実質的な誤差算出基準値をそれぞれＲ（０），Ｒ（１），Ｒ（２），Ｒ（３）と表してある。変動閾値マトリクス値が「１」の場合には、異なる値の変動閾値Ｖ１，Ｖ２，Ｖ３を用いた四値化処理が行われるので、それに応じて、４つの誤差算出基準値Ｒ（０），Ｒ（１），Ｒ（２），Ｒ（３）が存在する。
【００８９】
図１７は、変動閾値マトリクス内において特定の値（この実施形態では「１」）の場合についてのみより高い量子化レベルで量子化すること（すなわち、この実施形態では四値化すること）の効果を説明するための図である。上述のとおり、図１４（ａ）に示す変動閾値マトリクスを適用することによって、量子化閾値を二次元方向に周期変動させることができ、結果として一定の線数（たとえば１４１線）の網点により写真画像の階調を表現することができる。この場合に、網点の核となるドットＤＣが形成された場合に、このドットのＤＣの周りの画素を四値化すると、網点を構成するドットのＤＣの膨らみ方または縮み方が、ドットＤＣの周囲の画素を二値化する場合に比較して、より微細に設定することができる。すなわち、たとえば、レーザ走査ユニットを含む出力部４において、１画素あたりのレーザ発光時間またはレーザ発光強度を四値（２ビット）で表された画像データに基づいて４段階に制御することにより、出力部４における記憶画素の大きさを４段階に制御することができる。
【００９０】
この実施形態では、量子化閾値の周期変動における鞍部（変動閾値マトリクスの「０」に対応する。）の近傍の画素（変動閾値マトリクスのマトリクス値「１」に該当する。）について、四値化処理を行うこととしている。これにより、ドットの膨らみ方まは縮み方の段階数が、量子化レベルを二値化にのみ固定している場合に比較して詳細に定めることができ、滑らかな階調表現が可能となる。
この実施形態では、「１」の画素に対して四値化処理を行っているが、これは一例にすぎず、たとえばマトリクス値「２」の画素に対して量子化レベルの高い量子化（この実施形態では四値化）処理を行ってもよいし、マトリクス値「１」および「２」の両方に関して、量子化レベルの高い量子化処理を行ってもよい。
【００９１】
以上のように、この実施形態によれば、注目画素を含む第１画像ブロック（１３×７画素のマトリクス）内の画素間の画像データの差の総和ＴＯＴＡＬ＿ＳＵＭと、第１画像ブロック内の画像データの平均値ＡＶに応じて定められた閾値ＴＨ（ＡＶ）との大小比較結果に基づいて、写真領域の画素をそれ以外の領域を構成する画素から分離して抽出することができる。
さらに、第２画像ブロック（５×５画素のマトリクス）内の画素における主走査方向の画像データの差の総和Ｈ＿ＳＵＭおよび副走査方向の画像データの差の総和Ｖ＿ＳＵＭを所定の閾値ＨＬ，ＨＨ，ＶＬ，ＶＨと比較することによって、細線領域を構成する画素が写真領域に属するものと誤判定されることを防止できる。したがって、極めて正確に写真画像に属する画素を抽出できる。
【００９２】
さらに、分離前フィルタ処理部３９によって平滑化処理が施された後の画像データに対して上述のような領域分離処理を行うことにより、高線数（１７５線以上）の網点領域を写真領域に属するものと確実に判定することができる。これによって、中間調処理部３７における網点化処理（改良型誤差拡散処理）によってモアレが生じるおそれのない高線数の網点画像については、写真画像と同様に良好な階調性での画像再生を行うことができる。
【００９３】
さらに、この実施形態では、写真領域に属すると判定された画素に対しては、二次元方向に周期変動する変動閾値Ｖ１，Ｖ２，Ｖ３を用いた改良型誤差拡散処理が施されることにより、１７５線以上の高線数の網点画像および写真画像に対して網点化処理を施すことができる。これによって、電子写真プロセスによる画像形成において有利なドット集中型中間調処理が可能となり、階調表現に優れた再生画像を得ることができる。
【００９４】
その一方で、写真領域に属しないと判定された画素に対しては、量子化閾値を一定値に固定した通常型誤差拡散処理が行われるので、文字画像、線画画像、網点画像（低線数のもの）については、良好な解像性で画像を再生することができるとともに、モアレ縞が発生することもない。
また、この実施形態の改良型誤差拡散処理では、量子化閾値の変動と同位相で発生誤差の算出基準値を変動させることとしているので、量子化閾値の変動に起因する大きな量子化誤差が注目画素の周辺の周辺画素に分配されることがない。したがって、網点を構成すべき黒画素を効果的に集中させることができるので、の改良型誤差拡散処理による網点化処理を良好に行うことができる。しかも、画像の濃度を確実に保存できるという利点もある。
【００９５】
さらには、この実施形態では、網点の核となるドットの周囲の画素について、二値化処理ではなく四値化処理を施すことによって、網点の膨らみ方または縮み方の段階数を増大させているので、高線数の網点画像および写真画像を滑らかな階調性で表現することが可能になる。
以上、この発明の一実施形態について説明したが、この発明は他の形態で実施することもできる。たとえば、上述の実施形態では、高い量子化レベル（四値化）が適用される閾値マトリクス値が固定されているけれども、注目画素を含む第３の画像ブロック（たとえば、変動閾値マトリクスと同程度の大きさのマトリクス領域）について、この第３の画像ブロック内の画像データの平均値を求め、この平均値に基づいて、高い量子化レベルの量子化処理（四値化処理）を施すべきマトリクス値を可変設定することとしてもよい。この場合、第３の画像ブロック内の画像データの平均値とマトリクス値との関係は、上記図１４（ａ）の変動閾値マトリクスを適用する場合には、下記表１のように設定してもよい。
【００９６】
【表１】

【００９７】
このようにすることによって、いずれの濃度領域においても、ドット（網点）の周縁部を構成する画素に対して、確実に４値化処理を施すことができるので、より良好な階調再現性を実現できる。
また、上述の実施形態では、量子化閾値の周期変動とともに周期変動する発生誤差算出基準値は、変動閾値の値に対して一定の振幅を有しているが、この振幅は、注目画素の近傍の画像データの大小に応じて変動することとしてもよい。たとえば、低濃度領域において当該振幅を大きくすることにより、ドットの集中を抑制することができ、ドットの集中による違和感を抑制して美しい再生画像を実現することができる。これに対して、高濃度領域では、当該振幅を小さく設定することにより、ドットを良好に集中させることができ、良好な階調再現性を実現できる。
【００９８】
また、上述の実施形態では、量子化レベルを二値化と四値化との二段階に切り換えているが、量子化レベルの切り換えは三段階以上に行われてもよい。たとえば、二値化処理によってドットの核が形成される画素の周囲の画素を三値化することとし、さらにその周囲の画素について四値化処理を行うようにしてもよい。
また、上述の実施形態では、黒画素（網点を形成するドットの核）の周囲の画素について四値化処理を行うようにしているが、白画素の周囲に四値化される画素を配置して、白画素の周囲にグレーの画素（小面積の黒画素）が形成されるようにしてもよい。すなわち、この場合には、量子化閾値の変動周期の頂点の近傍の画素に関して高い量子化レベルでの量子化処理が行われることになる。
【００９９】
さらに、上記の実施形態では、この発明がディジタル複写機に適用された例について説明したが、この発明は、画像データを処理する装置に対して広く適用することができ、たとえばファクシミリ装置や、入力画像データに画像処理を施して記録シート上に画像を形成するプリンタに適用することもできる。
その他、特許請求の範囲に記載された事項の範囲で種々の設計変更を施すことが可能である。
【図面の簡単な説明】
【図１】この発明の一実施形態に係る画像処理装置が適用された画像形成装置の一例であるディジタル複写機の電気的構成を示すブロック図である。
【図２】画像処理部の機能的な構成を説明するためのブロック図である。
【図３】領域分離機能部の処理内容を説明するためのフローチャートである。
【図４】写真領域の画素の抽出のために用いられる第１画像ブロック（１３×７画素のマトリクス）の例を示す図である。
【図５】細線領域の画素の抽出のために用いられる第２画像ブロック（５×５画素のマトリクス）の例を示す図である。
【図６】第１画像ブロック内の画素の画像データの平均値ＡＶに対応した写真領域判定のための閾値ＴＨ（ＡＶ）の設定例を説明するためのグラフである。
【図７】第２画像ブロックにおける主走査方向濃度差総和Ｈ＿ＳＵＭと副走査方向濃度差総和Ｖ＿ＳＵＭとの関係の分布例を示す図である。
【図８】分離前フィルタ処理部の構成を説明するための図である。
【図９】分離前フィルタ処理部による処理内容を説明するための図である。
【図１０】第１画像ブロックのマトリクスの大きさを説明するための図である。
【図１１】中間調処理部の機能的な構成を説明するためのブロック図である。
【図１２】量子化誤差の累積および分配を説明するための図である。
【図１３】パラメータ生成処理部の機能的な構成を説明するためのブロック図である。
【図１４】変動閾値マトリクスメモリに格納された変動閾値マトリクス、およびこの変動閾値マトリクスの各マトリクス位置に対応した変動閾値マトリクスメモリのアドレスを示す図である。
【図１５】変動閾値テーブルＴ１，Ｔ２，Ｔ３の設定例を示す図である。
【図１６】量子化閾値の周期変動とそれに応じた誤差算出基準値の周期変動とを示す図である。
【図１７】変動閾値マトリクス内における特定の値についてより高い量子化レベルで量子化することの効果を説明するための図である。
【図１８】量子化閾値の変動の影響による発生誤差（量子化誤差）の変動を説明するための図である。
【符号の説明】
１ 読取部
３ 画像処理部
４ 出力部
３７ 中間調処理部
３８ 領域分離機能部
３９ 分離前フィルタ処理部
４０ 判定閾値テーブル
５１ パラメータ生成処理部
５２ 誤差拡散処理部
６１ 誤差加算処理部
６２ 累積誤差メモリ
６３ 量子化処理部
６４ 誤差演算部
６５ 誤差補正部
６６ 誤差分配処理部
７１ 固定閾値設定部
７２ 変動閾値設定部
７３ 閾値選択部
７４ 補正値演算部
８０ 変動閾値マトリクスメモリ
８１ 変動閾値メモリ
８２ 変動閾値メモリ
８３ 変動閾値メモリ
８４ 変動閾値選択部
８５ 変動閾値選択情報生成部
８６ 変動閾値生成部
Ｔ１ 変動閾値テーブル
Ｔ２ 変動閾値テーブル
Ｔ３ 変動閾値テーブル[0001]
TECHNICAL FIELD OF THE INVENTION
The present invention relates to an image processing device and an image processing method for performing a so-called error diffusion process.
[0002]
[Prior art]
2. Description of the Related Art In an image forming apparatus such as a digital copying machine, a printer, and a facsimile, a so-called pseudo halftone process is performed when converting multivalued image data representing density gradation into binary image data. In particular, since multi-valued image data representing a photographic image contains many halftones, pseudo halftone processing is indispensable. One of such pseudo halftone processes is an error diffusion process.
[0003]
In the error diffusion process, an error that occurs when converting the multi-valued image data of the target pixel into binary image data using a predetermined threshold value is set to a peripheral pixel around the target pixel that has not been binarized yet. The process of distributing and changing the multi-valued image data of the peripheral pixels is repeatedly executed. As a result, the density of dots output by the image forming unit is controlled, and an output image expressing a pseudo-halftone is obtained. An image forming unit that forms an image on a recording sheet by an electrophotographic process is often used as the image forming unit.
[0004]
For example, a case will be described in which pixels having image data (white: 0, black: 255) expressing multi-level density of 256 gradations are binarized by the above-described error diffusion processing (binary error diffusion processing).
In the error diffusion process, the image data of the pixel of interest, which is compared with the binarization threshold, is added to the image data representing the original multi-value density of the pixel, which is the cumulative value of the generated error distributed from other pixels. The error has been added.
[0005]
If the binarization threshold used for binarization is an intermediate value “127” of 256 gradations, it is determined whether the binary density of the target pixel is white (0) or black (1). Is performed according to the following criteria (A) and (B).
(A) When target pixel value + cumulative error> 127,
The binary density of the target pixel is set to black (1).
(B) When target pixel value + cumulative error ≦ 127,
The binary density of the target pixel is set to white (0).
[0006]
The error generated at the time of the determination based on the criteria (A) and (B) is distributed to peripheral pixels that are located around the target pixel and have not been binarized at that time. The error generated when the pixel value of interest satisfies one of the above criteria (A) and (B) is such that black (1) and white (0) of binary gray scale correspond to 255 and 0 in 256 gray scales, respectively. Thus, they are obtained by the following equations (A1) and (B1).
(A1) Generation error = (cumulative error + target pixel value) -255
(B1) Generated error = (cumulative error + target pixel value) −0
The error thus calculated is distributed to peripheral pixels according to a predetermined diffusion ratio.
[0007]
However, in such general error diffusion processing, dots are randomly dispersed, so that many isolated dots are likely to occur. However, in the formation of an image by an electrophotographic process, the size of each dot is not always stable, so that the size of an isolated dot cannot be accurately controlled. Therefore, when the general error diffusion processing is applied to an electrophotographic process, the stability of gradation expression is insufficient, and is not necessarily suitable for the expression of a photographic image.
[0008]
Another example of halftone processing is organized dithering. The systematic dither processing is a dot dispersion type (frequency modulation method) in which dots of the same size are densely and densely arranged according to the gradation, and the diameter of each dot is adjusted while maintaining the center interval of the dots constant. (Amplitude modulation method) which increases or decreases according to the dot concentration. Of these, from the viewpoint of gradation stability, it is said that the dot concentration type is more suitable for expressing halftones by an electrophotographic process.
[0009]
[Problems to be solved by the invention]
In view of the above, the applicant of the present application has disclosed in Japanese Patent Application No. 11-352825, which has previously filed, an improved type in which the threshold value at the time of the binarization process is periodically changed in the two-dimensional direction to increase the degree of concentration of dots. An error diffusion process was proposed. In the improved error diffusion process, dots are likely to be periodically concentrated, so that a so-called halftone dot process can be performed. Moreover, in the systematic dither processing, there is a problem in the resolution of a character image and a line drawing image, whereas in the improved error diffusion processing, the resolution of a character image and a line drawing image can be satisfied.
[0010]
Therefore, by using the above-described improved error diffusion processing for the pseudo halftone processing, it is not necessary to change the processing according to which type of image area the pixel belongs to, and the image processing is significantly simplified. Is obtained.
However, the improved error diffusion process has a problem that a large binarization error occurs when the threshold value is largely changed from the intermediate value to the white or black side. That is, as shown in FIG. 18A, if the binarization threshold is fixed at the intermediate value “127”, the maximum occurrence error is 127. On the other hand, as shown in FIG. 18B, for example, when the binarization threshold is greatly changed to the black side, a binarization error larger than “127” occurs.
[0011]
This large error is distributed to peripheral pixels. Since the peripheral pixels to which the large error has been distributed have a large image data value, there is a high possibility that the peripheral pixels are binarized into black pixels.
Therefore, in the above-described prior art, since the occurrence error varies in accordance with the variation of the threshold value, the probability that white pixels and black pixels alternately occur increases. As a result, the purpose of concentrating dots could not be effectively achieved.
[0012]
An object of the present invention is to provide an image processing apparatus and an image processing method that can effectively concentrate dots by suppressing a variation in an occurrence error caused by a variation in a quantization threshold, thereby achieving excellent tone reproducibility. It is to provide a processing method.
[0013]
Means for Solving the Problems and Effects of the Invention
According to the first aspect of the present invention, there is provided an image processing method for quantizing image data representing density gradation of pixels constituting an image into binary or more discrete value data smaller than a maximum density gradation value. A quantization unit (63) for quantizing the image data of the pixel of interest using a quantization threshold, and a quantization by the quantization unit.Threshold value setting means (51, 72) for periodically changing the threshold value, and a deviation of a quantization threshold value set by the threshold value setting means from an intermediate gradation value which divides the entire gradation range into two, is corrected by an error correction. A correction value calculating means (74) for calculating a value, and an amountTone value corresponding to discrete value data after child conversionAnd by reducing the generated error correction value calculated by the correction value calculating means, the generated error which is a quantization error generated at the time of quantization by the quantizing means is reduced.Means for calculating an occurrence error (64), 65), And the error diffusion means (61, 61) which distributes the generated error calculated by the generated error calculation means to the peripheral pixels around the pixel of interest and which have not been quantized yet, and changes the image data of the peripheral pixels. 62, 66)AndAn image processing apparatus characterized by including: It should be noted that the alphanumeric characters in parentheses indicate corresponding components and the like in embodiments described later. Hereinafter, the same applies in this section.
[0014]
According to the invention,amountSince it is possible to suppress the fluctuation of the error caused by the periodic fluctuation of the quantization threshold, it is possible to enhance the dot concentration effect due to the periodic fluctuation of the quantization threshold. As a result, an error diffusion process excellent in gradation expression can be realized, and for example, a photographic image or the like can be reproduced with good gradation characteristics.
[0016]
The invention according to claim 2 isThe threshold value setting means sets a plurality of quantization threshold values, and the correction value calculation means calculates a deviation of an average value of the plurality of quantization threshold values set by the threshold value setting means from the intermediate gradation value. Calculate as an error correction valueThe image processing apparatus according to claim 1, wherein.
[0017]
Claim3According to the invention described above, the threshold setting unit fixes or periodically varies the quantization threshold in the quantization unit in accordance with a region separation signal indicating which type of image region the pixel of interest belongs to. 2. The method according to claim 1, wherein the setting is switched.Or 2An image processing apparatus as described in the above.
[0018]
Claim4The described invention is an image processing method for quantizing image data representing the density gradation of pixels constituting an image into binary or more discrete value data smaller than the maximum density gradation value, A quantization step (63) of quantizing using a quantization threshold;A threshold setting step (51, 72) for setting the quantization threshold by changing the cycle periodically; and a deviation of the quantization threshold set by the threshold setting step from an intermediate gradation value that divides the entire gradation range into two. A correction value calculation step (74) for calculating as an occurrence error correction value,Tone value corresponding to discrete value data after child conversionAnd the generated error correction value calculated in the correction value calculation step is reduced to reduce the generated error, which is a quantization error generated at the time of quantization in the quantization step.Calculating occurrence error calculation step (64, 65) And an error diffusion step (61, 62, 66) of distributing the calculated error to peripheral pixels which are not quantized yet around the target pixel and change the image data of the peripheral pixels.AndAn image processing method comprising: This method corresponds to the first aspect of the present invention.
[0019]
Claim5According to the invention described above, the threshold setting step is to fix or periodically change the quantization threshold in the quantization step in accordance with an area separation signal indicating which type of image area the target pixel belongs to. Switching the setting or not4It is an image processing method described. This method corresponds to the fourth aspect of the present invention.
In addition,Claims 4 and 5Regarding the invention of the image processing method,,Request2Modifications similar to those described in are also possible.
[0020]
BEST MODE FOR CARRYING OUT THE INVENTION
Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings.
FIG. 1 is a block diagram showing an electric configuration of a digital copying machine as an example of an image forming apparatus to which an image processing apparatus according to an embodiment of the present invention is applied. This digital copying machine has a reading unit 1 provided with a photoelectric conversion element (for example, a line sensor such as a CCD) for reading an optical image of a document and converting it into a corresponding electric signal.
[0021]
The electric signal output from the reading unit 1 is input to an analog circuit unit 2 having an automatic gain control (AGC) circuit 21 and an analog / digital (A / D) converter 22. The AGC circuit 21 has a function of amplifying a minute analog electric signal from the reading unit 1 to an A / D conversion range (reference voltage) in the A / D converter 22. The A / D converter 22 quantizes the electric signal amplified by the AGC circuit 21 to generate a digital image signal. For example, a signal from the reading unit 1 having a minute amplitude of 1 volt is amplified by the AGC circuit 21 into an electric signal having an amplitude of 5 volts, and the 8-bit A / D converter 22 performs 0 (00h) to 255 (FFh). ), Digital image data expressing the density gradation in 256 levels is generated.
[0022]
The image data is given to the output unit 4 after undergoing various image processings by the image processing unit 3. The output unit 4 includes, for example, a laser scanning unit. That is, this digital copying machine has an image forming section for forming a copy image of a document according to an electrophotographic process. The image forming section includes a photoconductor, the above-described laser scanning unit for writing an electrostatic latent image corresponding to a copy image of a document on the photoconductor, and a developing device, a transfer unit, and a cleaning unit disposed around the photoconductor. And The electrostatic latent image written on the photoconductor by the laser scanning unit is developed into a toner image by a developing device. This toner image is transferred to the surface of a recording sheet such as a paper sheet or an OHP sheet (transparent sheet) in the transfer unit. The transferred toner image is heated and pressed by a fixing unit to be fixed on the recording sheet.
[0023]
The reading unit 1 includes a light source for illuminating a document, and an optical system for forming reflected light from the document on a detection surface of a line sensor. The reading unit 1 performs main scanning of the document by electrical scanning in the line sensor. The document is sub-scanned by relative movement of the line sensor and the optical system with respect to the document. Of course, the light source and the optical system may be moved, and the sub-scanning of the document may be achieved by moving the document while keeping the light source and the optical system stationary.
[0024]
FIG. 2 is a block diagram for explaining a functional configuration of the image processing unit 3. The image processing unit 3 includes a shading correction processing unit 31, an input γ correction processing unit 32, a black-and-white inversion processing unit 33, a zoom processing unit 34, a filter processing unit 35, an output γ correction processing unit 36, and a halftone processing unit 37. Each of these processing units sequentially processes input image data.
Further, the image processing unit 3 is supplied with an area separating function unit 38 for determining whether or not a pixel corresponding to the input image data is a pixel forming a photographic area, and to the area separating function unit 38. A pre-separation filter processing unit 39 for performing pre-processing on the image data is provided. The result of determination by the region separation function unit 38 is input to the filter processing unit 35, the output γ correction processing unit 36, and the halftone processing unit 37. In the filter processing unit 35, the output γ correction processing unit 36, and the halftone processing unit 37, the processing contents are changed depending on whether or not the pixels corresponding to the input image data form a photographic area.
[0025]
The analog signal output from the reading unit 1 is not uniform in the main scanning direction even when a blank image is read. This is because the light source and the optical system have different light distribution characteristics at the center and the end in the main scanning direction, and there is a variation in sensitivity among the pixels read by the line sensor. The shading correction processing unit 31 corrects such variations in image data between pixels due to the characteristics of the reading unit 1.
[0026]
The image data that has been processed by the shading correction processing unit 31 is subjected to intensity-density conversion by the input γ correction processing unit 32. The input γ correction processing unit 32 generates image data having gradation characteristics proportional to the density of the document by correcting the reading characteristics of the reading unit 1.
The black-and-white inversion processor 33 inverts the black-and-white logic of the input image data. The reading unit 1 outputs a high-density portion (black area) of a document at a low voltage. That is, the higher the density of the document image, the lower the output voltage of the reading unit 1. The black-and-white inversion processing unit 33 converts the image data by inverting the logic of the image data so that the value of the image data increases as the density of the image increases.
[0027]
The zoom processing unit 34 is for realizing a zooming function by digital processing.
The filter processing unit 35 performs edge enhancement processing or smoothing processing on image data. In this embodiment, an area separation signal from the area separation function section 38 is input to the filter processing section 35. In response to the region separation signal, the filter processing unit 35 performs a smoothing process on the pixels belonging to the photographic region, and performs an edge enhancement process on the pixels constituting the other regions. As a result, the photographic area can be smoothly reproduced, and the character area, the line drawing area, and the halftone dot area having a low number of lines can be reproduced with good resolution.
[0028]
The output γ correction processing unit 36 performs processing for correcting γ characteristics of the output engine in the output unit 4 on the image data. That is, in the electrophotographic process, it is difficult to obtain an image density output that changes linearly with respect to input image data due to the charging characteristics of the photosensitive member and the toner. By correcting the γ characteristic of such an output engine, an image density output that changes almost linearly with respect to the input image data is realized.
The halftone processing unit 37 is for realizing a so-called pseudo halftone process by converting (quantizing) the image data in which the density is expressed by 256 gradations into a discrete value of two or more and less than 256 values. is there. In this embodiment, the halftone processing unit 37 is configured to perform binarization processing in principle and to perform quaternization processing as needed.
[0029]
More specifically, the halftone processing unit 37 includes a normal error diffusion process in which the quantization threshold is fixed to a fixed value, and an improved error diffusion process in which the quantization threshold is periodically changed in two dimensions (dotting process, (Dot-concentration type halftone processing). In the improved error diffusion processing, based on the binary error diffusion processing, quaternization processing (quaternary error diffusion processing) is performed on pixels near the dot concentration position, and the halftone dot swelling method is performed. Or the number of steps of the contraction method is increased.
[0030]
Switching between the normal type error diffusion process and the improved type error diffusion process in the halftone processing unit 37 is performed in accordance with a segmentation signal from the segmentation function unit 38. That is, the improved error diffusion processing is performed on the image data of the pixels belonging to the photographic area, and the normal error diffusion processing is performed on the image data of the pixels forming the other areas.
FIG. 3 is a flowchart for explaining the processing contents of the area separation function unit 38. The area separation function unit 38 sequentially sets individual pixels constituting the read document image as a target pixel in accordance with the reading order in the reading unit 1 and determines whether or not the target pixel belongs to a photograph area. For that purpose, first, the average value of the image data of the pixels in the first image block of a certain size including the target pixel is calculated (step S1).
[0031]
For example, as shown in FIG. 4, the first image block is composed of pixels A1 to A13, B1 to B13, C1 to C13, D1 to D13, E1 to E13, F1 to F13, G1 to It may be composed of a matrix of 13 × 7 pixels composed of G13 and centered on the target pixel D7. In this case, the average value AV of the image data of the pixels in the first image block composed of the matrix of 13 × 7 pixels is calculated according to the following equation (1).
[0032]

The segmentation function unit 38 further obtains a difference in image data between adjacent pixels with respect to the first image block, that is, a pixel in the matrix of 13 × 7 pixels, and obtains a difference between image data of all pixels in the matrix. The sum of the differences is calculated (step S2).
[0033]
Specifically, the area separation function unit 38 calculates the total sum TOTAL_DH of the image data differences between pixels in the main scanning direction in the reading unit 1 according to the following equation (2). In addition, the total TOTAL_DV of the difference of the image data between the pixels adjacent in the sub-scanning direction in the reading unit 1 is calculated according to the following equation (3). Then, by adding these, the total TOTAL_SUM of the differences between the image data in the matrix of 13 × 7 pixels is calculated (see the following equation (4)). Of course, the total TOTAL_SUM of the differences between the pixel data may be directly obtained without calculating TOTAL_DH and TOTAL_DV.
[0034]

The region separating function unit 38 also includes a horizontal thin line region in which horizontal lines in the main scanning direction in the reading unit 1 are formed at high density in parallel, and a vertical line in the sub-scanning direction in the reading unit 1 is high in parallel. Formed with densityVerticalIn order to distinguish the thin line region from the photograph region, it is determined whether the pixel of interest belongs to the thin line region (horizontal thin line region or vertical thin line region).
[0035]
For this determination, the area separation function unit 38 determines, for each pixel in the second image block having a certain size including the pixel of interest, the image data between pixels adjacent in the first direction (for example, the main scanning direction). The difference is obtained, and the difference in the image data is accumulated for all the pixels in the second image block to obtain the sum of the density differences in the first direction. Further, the area separating function unit 38 obtains, for each pixel in the second image block, a difference in image data between pixels adjacent in a second direction (for example, a sub-scanning direction) orthogonal to the first direction, and By summing up the data differences for all the pixels in the second block, the second direction density difference sum is obtained.
[0036]
Specifically, as shown in FIG. 5, pixels B5 to B9, C5 to C9, D5 to D9, E5 to E9, and F5 to F9 centering on the pixel of interest D7 using an appropriate buffer memory. Is formed as a second image block. Then, based on the following equation (5), the main scanning direction density difference sum H_SUM is calculated as the first direction density difference sum, and based on the following equation (6), the sub-scanning direction density sum as the second direction density difference sum is calculated. The difference sum V_SUM is calculated (step S3 in FIG. 3).
[0037]

Next, the area separation function unit 38 refers to the determination threshold table 40 (see FIG. 2) to determine the determination threshold TH (AV) (photo area) as a function of the average value AV obtained in step S1. (Determination threshold). The read determination threshold TH (AV) is compared with the total sum TOTAL_SUM of the differences between the image data in the matrix of 13 × 7 pixels (step S4). That is, if the total sum TOTAL_SUM of the differences between the image data is equal to or greater than the determination threshold TH (AV) (NO in step S4), it is determined that the pixel of interest is a pixel that forms an image of a type other than the photographic region ( Step S7). Therefore, the area separation function unit 38 outputs a corresponding area separation signal (for example, a 1-bit signal whose one value is set to “0”).
[0038]
On the other hand, if the sum TOTAL_SUM of the differences between the image data is less than the determination threshold TH (AV) (YES in step S4), a tentative determination is made that the target pixel D7 belongs to the photographic area. However, whether or not this determination is correct is determined in step S5 described later.SThe determination is made again by the processing of step 6.
FIG. 6 is a graph for explaining a setting example of the threshold value TH (AV) corresponding to the average value AV. FIG. 6 shows the relationship between the average value AV and the total sum TOTAL_SUM of the differences between the image data. A setting example of the threshold value TH (AV) with respect to the average value AV is represented by a solid curve. Further, the distribution of the relationship between the average value AV and the total sum TOTAL_SUM of the image data in the photographic region is represented by a symbol “○”, and the distribution is other than the photographic region (the character region, the line drawing region, and the low-frequency halftone screen). The distribution of the relationship between the average value AV and the total sum TOTAL_SUM of the image data in the point area) is represented by a symbol “x”.
[0039]
As understood from FIG. 6, the threshold value TH (AV) is set so as to be located between the distribution area of the symbol “シ ン ボ ル” and the distribution area of the symbol “×”. Therefore, since the total sum TOTAL_SUM of the image data is less than the threshold value TH (AV), it is possible to determine that the target pixel belongs to the photographic area, and the total TOTAL_SUM of the difference between the image data is determined according to the average value AV. If it is equal to or greater than the set threshold value TH (AV), it can be determined that the target pixel belongs to an area other than the photograph area.
[0040]
Referring again to FIG. 3, the area separating function unit 38 further determines the magnitude of the sum of the density differences H_SUM in the main scanning direction and the sum V_SUM of the density differences in the sub-scanning direction obtained for the matrix of 5 × 5 pixels constituting the second image block. The relationship is checked by the processing of steps S5 and S6, and it is determined based on the result whether or not the target pixel D7 belongs to the thin line region. If it is determined that the target pixel D7 belongs to the thin line region, it is determined that the target pixel D7 does not belong to the photographic region by the processing in step S7, and a region separation signal “0” corresponding to the target pixel D7 is generated.
[0041]
More specifically, as shown in FIG. 7, two threshold values HL and HH (HH> HL) are set for the horizontal density difference sum H_SUM, and similarly, for the sub-scanning direction density difference sum V_SUM, Thresholds VL and VH (VH> VL) are set. For example, thresholds HL, HH; VL, VH are set to the following values.
HL = 50 HH = 150
VL = 50 VH = 100
As shown in FIG. 7, in the plane in which the sum of the density differences V_SUM in the sub-scanning direction is set on the horizontal axis and the sum of the density differences H_SUM in the main scanning direction is set on the vertical axis, fine lines along the horizontal direction (main scanning direction) are dense. As for the pixels constituting the horizontal thin line region, many pixels are distributed in an area HF where the sum of the density differences V_SUM in the sub-scanning direction is large and the sum of the density differences H_SUM in the main scanning direction is small. In addition, for pixels belonging to a vertical thin line region in which fine lines along the vertical direction (sub-scanning direction) are dense, many pixels are arranged in an area VF where the total sum of density differences V_SUM in the sub-scanning direction is small and the total sum H_SUM in the main scanning direction is large. Are distributed. Then, with respect to the halftone image and the original photographic image that should be regarded as the photographic region, the pixel distribution is obtained in the region PS near the center of the V_SUM-H_SUM plane.
[0042]
Therefore, in this embodiment, the region separation function unit 38 determines that the sum of the density differences H_SUM in the main scanning direction for the target pixel is equal to or greater than the threshold value HH and the sum of the density differences in the sub-scanning direction is V_SUM is equal to or less than the threshold value VL. (YES in step S5 in FIG. 3), it is determined that the target pixel belongs to the thin line area (in this case, the vertical thin line area). That is, it is determined that the target pixel does not belong to the photographic region, and a region separation signal “0” indicating this is generated (step S7).
[0043]
If the sum H_SUM of the density differences in the main scanning direction for the pixel of interest is equal to or smaller than the threshold HL and the sum V_SUM of the density differences in the sub-scanning direction is equal to or larger than the threshold VH (YES in step S6), It is determined that the target pixel belongs to the thin line region (in this case, the horizontal thin line region). That is, it is determined that the pixel of interest does not belong to the photographic area, and an area separation signal “0” corresponding to the pixel is generated (step S7).
If the determinations in steps S5 and S6 are both negative, that is, if the total density difference H_SUM in the main scanning direction is equal to or greater than the threshold value HH, but the total density difference V_SUM in the sub-scanning direction exceeds the threshold value VL, When the total density difference H_SUM is equal to or less than the threshold value HL, but the total density difference V_SUM in the sub-scanning direction is less than the threshold value VH, and when the total density difference H_SUM in the main scanning direction is in the range of HL to HH, the pixel of interest D7 Is determined not to belong to the thin line region. That is, the target pixel D7 is re-determined as belonging to the photographic area (step S8). In this case, the area separation function unit 38 sets, for example, a 1-bit area separation signal to a value (“1” in this embodiment) corresponding to the photograph area.
[0044]
In this way, the determination based on the sum of the density differences between the pixels in the matrix of 13 × 7 pixels (YES in step S4) is confirmed again.
FIG. 8 is a diagram for explaining the process of the pre-separation filter processing unit 39. The pre-separation filter processing unit 39 basically constitutes integration processing means for performing processing as an integration filter. That is, as shown in FIG. 8A, a rhombus (in the main scanning direction) composed of a total of 13 pixels (a3, b2 to b4, c1 to c5, d2 to d4, and e3) centered on the target pixel c3. The image data of the pixel of interest c3 is smoothed using the image data of the pixels in the matrix of (square inclined at 45 degrees with respect to the pixel).
[0045]
As shown in FIG. 8B, the coefficients s1 to s6 multiplied by the image data of each pixel in the matrix of FIG. 8A are, for example, in the main scanning direction and the sub-scanning direction with respect to the target pixel c3. It is set to be almost symmetric. s1 to s6value is, May be set equal as in the following equation, or may be set to different values. That is, for example, the values of s1 to s6 may be determined so that the weights are different depending on the distance from the target pixel.
[0046]
s1 = s2 = s3 = s4 = s5 = s6 = 1
When s1 to s6 are set to the values of the above expression, the image data of the target pixel c3 is converted to the next value F (c3).

In this embodiment, for example, halftone dots having a screen angle of 45 degrees and the number of lines of 141 are formed in the improved error diffusion processing performed on the pixels belonging to the photographic area in the halftone processing unit 37. On the other hand, the reading resolution in the reading unit 1 is, for example, 600 dpi (dot per inch). Then, in the halftone dot of the 141 line, as shown in FIG. 9, the halftone dot is located at a position corresponding to the pixels p1 and p2 on the oblique 45-degree line separated by three pixels in the main scanning direction and the sub-scanning direction. Will exist.
[0047]
In general, the dots to be subjected to the halftone dot processing are better stabilized as the number of lines of the halftone dot in the intermediate processing is closer to the number of halftone dots. This is because the number of lines to be halftoned in the image output device is selected as high as possible within the range that can maintain the image reproducibility in the device, so if the number of lines to be halftoned is greater than that, This is because the stability of the reproduced dots deteriorates even when the dots are formed. That is, a stable dot cannot be formed even if a halftone dot having a line number equal to or larger than the halftone frequency in the halftone processing of the image output device is faithfully reproduced. Although it depends on the image output device and the number of lines for halftone processing, recently, processing is usually performed at 600 dpi. In general, the difference between the number of lines for halftone processing and the number of fine lines for a document is ± 50 lines. Within this range, the dot stability is maintained.
[0048]
On the other hand, from the viewpoint of preventing moiré fringes, it is preferable that the number of halftone dots in halftone processing and the number of halftone dots of a document to be image-processed are not close to each other. This is because the closer the number of lines is, the more moire fringes are generated that are noticeable to humans. Generally, Moire fringes hardly occur if they are separated by ± 50 lines of the number of lines in the halftone processing.
Since the halftone processing is usually performed on the high frequency side such as a photograph, the number of halftone processing lines (the number of halftone processing lines) and the number of lines + 50 lines , The boundary line number (for example, 175 lines) of the region separation in the halftone processing is set, the line number region higher than the line number is determined as the photograph region, and the line region below the line number is excluded as other than the photograph region. .
[0049]
The halftone frequency of the original is close to the halftone frequency of the halftone dot formed by the improved error diffusion processing unit by the halftone processing unit 37, and the difference in the number of halftone lines is a small value within the range of ± 50 lines. Sometimes, so-called moire fringes may be formed in the output image. Therefore, it is appropriate to perform not the improved error diffusion processing but the normal error diffusion processing in which the quantization threshold is fixed to a constant value, for the original pixels in such a line frequency range. On the other hand, a halftone dot image having a higher number of lines than a boundary number defined between 141 lines and 141 + 50 lines, which is a halftone line number formed by the improved error diffusion processing, is regarded as a photographic image and is improved. Good tone reproduction can be realized by applying the mold error diffusion process.
[0050]
Therefore, a boundary line number is set between a predetermined line number (141 lines in this embodiment) for performing the improved error diffusion processing and a line number +50 lines more than this (191 lines in this embodiment), and the boundary is set. An image having the number of lines equal to or greater than the number of lines is determined as a photographic region, and the other images are separated from the other regions. The number of boundaries is determined by the configuration of the pre-separation filter processing unit 39. Specifically, as the range of the matrix in FIG. 8A is increased, the number of lines causing interference decreases and the number of boundary lines decreases. Further, when the coefficients s1 to s6 of the matrix are different (for example, the center is higher), the number of boundary lines increases. In the case of this embodiment in which the coefficients s1 to s6 are determined equally using the matrix of FIG. 8A, the number of boundary lines is about 175 lines.
[0051]
As shown in FIG. 9, the above-described integration filter in the pre-separation filter processing unit 39 does not cause interference between halftone dots in the image data after the integration filter processing for the 141-line halftone document. (In FIG. 9, the halftone dots after the smoothing process are indicated by oblique lines.) When the input image is a halftone image having a high screen ruling of 175 lines or more, this integration filter causes good interference between adjacent halftone dots. As a result, the halftone image having a high screen ruling of 175 lines or more has a relatively small total TOTAL_SUM of the image data differences between pixels in the matrix of 13 × 7 pixels. It is determined that the pixels forming the halftone pixel having such a high screen ruling belong to the photographic area.
[0052]
Note that the first image block need not be a matrix of 13 × 7 pixels, and similarly, the second image block need not be a matrix of 5 × 5 pixels. Absent.
That is, the size of the first image block is determined by the distribution of the relationship between the above-described average value AV and the total sum TOTAL_SUM of the image data in the first image block, as shown in FIG. May be set to appear in the separated area.
[0053]
The reason why the matrix of 13 × 7 pixels is adopted in the above-described embodiment is to refer to an area which always includes a halftone dot forming a halftone image of 65 lines in a block. The 65-line halftone image is an image used for printing a photograph of a newspaper, for example, and is the minimum halftone frequency used in a general printed matter. Therefore, if it is possible to separate a halftone area of 65 lines or more as a halftone area (other than a photographic area), it is practically sufficient.
[0054]
For this purpose, at least one halftone dot constituting the halftone image of 65 lines must exist in the reference range (first image block). In this case, as shown in FIG. 10, if the halftone dot D is formed at a screen angle of 45 degrees, the interval between adjacent halftone dots along the main scanning direction is √2 / 65 inch. The interval between the halftone dots D along the sub-scanning direction is 1 / (65√2) inches.
Assuming that the reading resolution in the reading unit 1 is 600 dpi, the interval between adjacent halftone dots D along the main scanning direction is 13 (≒ 600 × √2 / 65) pixels, and the interval between halftone dots D along the sub-scanning direction. Is 7 (≒ 6.53 = 600 / (65 · √2)) pixels. Therefore, as described in the above-described embodiment, by setting the size of the first image block to a matrix of 13 × 7 pixels, a halftone dot region having 65 lines or more is separated as a region other than the photograph region. can do.
[0055]
As described above, it is understood that the size of the first image block depends on the halftone frequency and the reading resolution of the reading unit 1, and a matrix of 13 × 7 pixels is merely an example.
In order to reduce the capacity of the buffer memory for forming the first image block, the first image block of the first image block must be so selected that one halftone dot of the halftone image having the lowest screen frequency exists in the reference matrix. It is preferable to determine the size. However, the size of the first image block may be set so that more than one (not necessarily an integer) halftone dots exist in the first image block. That is, when it is intended to determine that a halftone dot area having 65 lines or more is not a photograph area, a matrix larger than a matrix of 13 × 7 pixels may be used as the first image block.
[0056]
The same applies to the above-described second image block, and the distribution of the relationship between the sum of the density differences H_SUM in the main scanning direction and the sum of the density differences V_SUM in the sub-scanning direction is regarded as a photograph as shown in FIG. The size of the pixel matrix may be determined so as to be separated from the power image and the thin line region image. Therefore, for example, the second image block may be a square pixel matrix smaller or larger than 5 × 5 pixels, or a rectangular or other matrix having different numbers of pixels in the main scanning direction and the sub-scanning direction. It may be. For example, a matrix of 13 × 7 pixels constituting the first image block can be directly used as the second image block, and TOTAL_DH and TOTAL_DV can be used instead of H_SUM and V_SUM, respectively.
[0057]
However, when the reference matrix is made large, there is a possibility that an error may occur in the area separation determination for the pixels that are initially processed in each line because there is little image data that can be referred to. From this point of view, it is preferable to adopt a matrix of a minimum necessary size for both the first image block and the second image block.
At both ends in the main scanning direction and both ends in the sub-scanning direction, the pixels constituting the first image block or the second image block are insufficient. For example, when processing the first pixel in each line as a target pixel, no pixel exists on the upstream side in the main scanning direction. In such a case, in the matrix of 13 × 7 pixels or the matrix of 5 × 5 pixels, the image data of the pixel at the non-existing matrix position is regarded as “0”, and the above-described region separation determination is performed.
[0058]
FIG. 11 is a block diagram for explaining the functional configuration of the halftone processing unit 37. The halftone processing unit 37 includes an area separation signal provided from the area separation function unit 38, target pixel coordinates indicating the coordinate position of the target pixel, and image data of the target pixel (in this embodiment, the image after the output γ correction processing). And a parameter generation processing unit 51 for generating a parameter for error diffusion processing based on the data. Further, the halftone processing unit 37 performs an error diffusion processing operation on the input image data based on the image data supplied from the output γ correction processing unit 36 and various parameters supplied from the parameter generation processing unit 51. A diffusion processing unit 52 is provided.
[0059]
The parameter generation processing unit 51 has thresholds V1, V2, V3 or F1, F2, F3 or F1, F2, F3 for quantizing input image data (binary or quaternary in this embodiment) by using a configuration to be described later. Is generated, and the generated error correction value EC for correcting an error (quantization error) generated at the time of quantization is generated and given to the error diffusion processing unit 52.
The error diffusion processing unit 52 includes an error addition processing unit 61 that adds a cumulative error stored in a cumulative error memory 62 to input image data. The accumulated error memory 62 accumulates and stores errors distributed to the pixel of interest from neighboring pixels for which quantization processing has already been completed. For example, as shown in FIG. 12 (a), for a target pixel X, a predetermined error diffusion coefficient (a) is determined from the surrounding quantized pixels a, b, c, d, e, and f. For example, the generation errors multiplied by 1/4 or 1/8) are accumulated and added. The accumulated error in this case is represented by the following equation, for example.
[0060]

The accumulated error is stored in the accumulated error memory 62 in association with each pixel. The accumulated error corresponding to the target pixel of the error diffusion processing is read out, and the error addition processing unit 61 converts the accumulated error into image data. Will be added.
The error diffusion processing unit 52 further converts the image data after the error addition processing into three quantization thresholds V1, provided respectively from the parameter generation processing unit 51, corresponding to the normal error diffusion processing and the improved error diffusion processing. A quantization processing unit 63 for performing quantization based on V2, V3 or F1, F2, F3 is provided. In this embodiment, the quantization processing unit 63 converts the image data into 2-bit data representing four values of 0, 1, 2, and 3.
[0061]
The quantization thresholds V1, V2, and V3 are fluctuation thresholds that are set to periodically fluctuate in a two-dimensional direction along the image plane, and a magnitude relationship of V1 ≦ V2 ≦ V3 is established between them. Further, the quantization thresholds F1, F2, and F3 are fixed thresholds, and a magnitude relationship of F1 ≦ F2 ≦ F3 (in this embodiment, particularly, F1 = F2 = F3) is established. The parameter generation processing unit 51 generates one of the variable thresholds V1, V2, V3 or the fixed thresholds F1, F2, F3. In response to this, the quantization processing unit 63 executes the next case determination I, II, III, and IV in this order based on the three quantization thresholds V1, V2, and V3 or F1, F2, and F3. , And quantizes (quaternizes) the image data to 0-3. In the process of executing the case determination in the order of I to IV, if any of the conditions is met, the subsequent determination is not performed and the quantized value (output data) is determined.
[0062]

[0063]
IV. When V1 or F1 ≧ image data after error addition processing,
The target pixel is quantized to “0”.
・ ・ ・ ・ ・ ・ (9)
In this way, when the quantization thresholds V1, V2, V3 or F1, F2, F3 have different values, the image data of the target pixel is quantized to four values of “0” to “3”. . If the thresholds V1, V2, V3 or F1, F2, F3 have the same value, the above-described case classification is performed.Fixed IBy sequentially executing steps IV to IV, the image data of the target pixel is binarized to “0” or “3”.
[0064]
An error generated as a result of the quantization processing by the quantization processing unit 63 is obtained by the error calculation unit 64. When the density gradation of the input image data is represented by 256 gradations of 0 to 255, the error calculation unit 64 calculates a generation error according to the following equation according to the output data (quantization result).

The output data “3” corresponds to a black pixel and corresponds to the maximum gradation value “255”. The output data “0” corresponds to a white pixel and corresponds to the minimum gradation value “0”. The output data "1" and "2" can be said to correspond to gray pixels, and correspond to the first halftone value "85" and the second halftone value "170", respectively. ing. However, here, the first and second halftone values are selected as halftone values that equally divide all the grayscale sections from 0 to 255 into four.
[0065]
The generated error calculated by the error calculation unit 64 is subjected to correction processing by the error correction unit 65. This correction processing is achieved by subtracting the generated error correction value EC given from the parameter generation processing unit 51 from the generated error obtained by the error calculation unit 64, as shown in the following equation (11). The generated error after the correction is subjected to distribution processing by the error distribution processing unit 66, and the accumulated error as a result of the distribution processing is stored in the accumulated error memory 62.
[0066]
Error after correction = Error in equation (10) −EC (11)
The processing in the error distribution processing unit 66 is the opposite of the processing of accumulating the errors shown in FIG. That is, the error generated in the target pixel X is multiplied by a predetermined error diffusion coefficient (1/4 or 1/8 in this example) as shown in FIG. , Are distributed to peripheral pixels that have not yet been binarized.
[0067]
FIG. 13 is a block diagram for explaining a functional configuration of the parameter generation processing unit 51. The parameter generation processing unit 51 calculates, on the image plane, a parameter corresponding to the normal type error diffusion processing in which the quantization threshold is fixed to a constant value, or a quantization threshold in the image plane based on the region separation signal provided from the region separation function unit 38. A parameter is generated for an improved error diffusion process that is set by dimensionally varying the period. That is, when the region separation signal has a value “1” indicating that the target pixel belongs to the photograph region, a parameter for the improved error diffusion process is generated, and the region separation signal indicates that the target pixel belongs to the photograph region. If the value is “0” indicating that the pixel belongs to a region other than the region, a parameter for normal error diffusion processing is generated.
[0068]
For such a function, the parameter generation processing unit 51 includes a fixed threshold value setting unit 71 that generates fixed threshold values F1, F2, and F3 for the normal error diffusion process, and a variable threshold value for the improved error diffusion process. A variable threshold value setting unit 72 for setting V1, V2, and V3; and selecting one of the fixed threshold values F1, F2, and F3 or the variable threshold values V1, V2, and V3 based on the region separation signal. A threshold value selection unit 73 to be provided and a correction value calculation unit 74 that calculates a generated error correction value EC for correcting a generated error and provides the calculated error correction value EC to the error diffusion processing unit 52 are provided.
[0069]
The correction value calculation unit 74 calculates an occurrence error correction value EC based on the region separation signal and the fluctuation threshold values V1, V2, and V3 given from the fluctuation threshold value setting unit 72. Specifically, when the region separation signal is “0” and the pixel of interest is a pixel belonging to a region other than the photographic region, the generated error correction value EC is set to a constant value (“0” in this embodiment). I do. On the other hand, when the value of the region separation signal is “1”, the generation error correction value EC is varied according to the values of the variation thresholds V1, V2, and V3 given from the variation threshold setting unit 72.
[0070]
In the present embodiment, the fixed threshold values F1, F2, and F3 generated by the fixed threshold value setting unit 71 are set to the same value (for example, “127” which is an intermediate value between 0 and 255). When the fixed threshold values F1, F2, and F3 are given to the error diffusion processing unit 52, the quantization processing unit 63 of the error diffusion processing unit 52 sets the input image data to any one of “0” and “3”. Binarize.
The fluctuation threshold value setting unit 72 includes a fluctuation threshold value matrix memory 80 storing a fluctuation threshold value matrix for setting a quantization threshold value that periodically fluctuates two-dimensionally in an image plane, A variation threshold value selection information generation unit 85 for finding which position in the matrix corresponds to the value of the variation threshold value matrix at that position as variation threshold value selection information, and the variation threshold value selection information and the image data value of the pixel of interest And a fluctuation threshold generation unit 86 that generates fluctuation thresholds V1, V2, and V3 based on the above.
[0071]
The fluctuation threshold generation unit 86 includes a fluctuation threshold memory 81 storing a fluctuation threshold table T1 for generating the fluctuation threshold V1, a fluctuation threshold memory 82 storing a fluctuation threshold table T2 for generating the fluctuation threshold V2. And a fluctuation threshold memory 83 storing a fluctuation threshold table T3 for generating the fluctuation threshold V3. Further, the variation threshold generation unit 86 selects the corresponding variation thresholds V1, V2, V3 from the variation threshold memories 81, 82, 83 based on the variation threshold selection information and the value of the image data of the pixel of interest, and selects the threshold. A variation threshold value selection unit 84 to be provided to the unit 73 is provided.
[0072]
When the segmentation signal has the value “1” corresponding to the photographic region, the threshold selection unit 73 selects the variation thresholds V1, V2, and V3 and supplies them to the error diffusion processing unit 52. If the segmentation signal is a value “0” indicating that the region is not a photographic region, the fixed thresholds F1, F2, and F3 set by the fixed threshold setting unit 71 are given to the error diffusion processing unit 52.
FIG. 14A is a diagram showing an example of the variation threshold matrix stored in the variation threshold matrix memory 80, and FIG. 14B is a variation threshold matrix memory 80 corresponding to each matrix position of the variation threshold matrix. FIG. FIG. 14A shows a variation threshold matrix considering the case where the image data is quaternized, and each matrix value is set to any one of 0, 1, 2, and 3. ing. A relatively small quantization threshold is set for a pixel corresponding to a small matrix value in the threshold matrix, and a relatively large quantization threshold is set for a pixel corresponding to a large matrix value. Therefore, if the matrix value corresponding to the pixel of interest is small, the pixel of interest is likely to be quantized to a high density pixel (black pixel), and if the matrix value corresponding to the pixel of interest is large, the pixel of interest is low. The possibility of being quantized into density pixels (white pixels) increases.
[0073]
The process of quantizing an image using the variation thresholds V1, V2, and V3 can be understood by imagining a situation in which the variation threshold matrix shown in FIG. 14A is spread two-dimensionally on the image plane. That is, in this case, the quantization threshold in the quantization processing unit 63 (see FIG. 11) periodically fluctuates in the main scanning direction and the sub-scanning direction (that is, two-dimensional direction). Therefore, dots (halftone dots) appear periodically in the image after quantization, and as a result, halftoning processing can be performed.
[0074]
In the example of FIG. 14A, a small matrix of 3 × 3 pixels in which a matrix value “0” (corresponding to a saddle point of a fluctuation period of a quantization threshold) is surrounded by a matrix value “1”, and a matrix value “3” (Corresponding to the apex of the fluctuation period of the quantization threshold) with a matrix of 3 × 3 pixels surrounded by a matrix value “2” are alternately arranged in the main scanning direction and the sub-scanning direction. As a result, a halftone dot having a screen angle of 45 degrees is formed. If the reading resolution of the reading unit 1 is 600 dpi, the halftone frequency is 141 lines.
[0075]
The variation threshold matrix in FIG. 14A is set to 6 × 6 pixels, but other matrix sizes such as 5 × 5 pixels and 7 × 7 pixels can be adopted. For example, if the matrix size of the matrix of n × n pixels is defined as “n”, the position in the variation threshold matrix corresponding to the pixel of interest, that is, the address of the variation threshold matrix memory 80, is the coordinates of the pixel of interest (main scanning coordinates and (Sub-scanning coordinates), and can be obtained by the following equation.
[0076]
Address = (main scanning coordinate 主 remainder of n + sub scanning coordinate ÷ remainder of n) × 8 (12)
This formula assumes that the maximum matrix size of the variation threshold matrix is “8” (see FIG. 14B), and the address of the variation threshold matrix memory 80 is specified by the variation threshold matrix memory. It will be changed appropriately according to the configuration of 80.
[0077]
The variation threshold selection information generating unit 85 reads the matrix value of the variation threshold matrix from the coordinates of the pixel of interest based on the above equation (12), and supplies the matrix value to the variation threshold selection unit 84 as variation threshold selection information. The variation threshold selection unit 84 reads out appropriate variation thresholds V1, V2, V3 from the variation threshold tables T1, T2, T3 of the variation threshold memories 81, 82, 83 based on the given variation threshold selection information.
FIG. 15 is a diagram illustrating a setting example of the variation threshold tables T1, T2, and T3. The fluctuation thresholds V1, V2, and V3 are set so that the relationship of V1 ≦ V2 ≦ V3 is satisfied.
[0078]
More specifically, in the present embodiment, the variation thresholds V1 (0), V2 (0), and V3 (0) corresponding to the matrix value “0” of the variation threshold matrix are, in principle, V1 (0) = V2 (0) = V3 (0), the variation thresholds V1 (0), V2 (0), V3 (0) are determined. Similarly, for the matrix values “2” and “3” of the variation threshold matrix, V1 (2) = V2 (2) = V3 (2) and V1 (3) = V2 (3) = V3 (3). The fluctuation thresholds V1, V2, and V3 are determined so that the relationship is established in principle.
[0079]
Then, as for the matrix value "1" which is a specific matrix value in the variation threshold matrix, in principle, V1 (1) <V2 (1) for the variation thresholds V1 (1), V2 (1), and V3 (1). ) <V3 (1) are established, so that the variation thresholds V1, V2, and V3 are determined.
Further, as for the variation thresholds V1, V2, and V3, a smaller variation threshold is assigned in principle as the matrix value of the variation threshold matrix is smaller. That is, the variation thresholds V1, V2, and V3 are determined so that the following relationship is established.
[0080]
V1 (0) <V1 (1) <V1 (2) <V1 (3)
V2 (0) <V2 (1) <V2 (2) <V2 (3)
V3 (0) <V3 (1) <V3 (2) <V3 (3) (12)
When the variation thresholds V1, V2, and V3 are given as quantization thresholds and the matrix value of the variation threshold matrix is 0, 2, or 3, the quantization processor 63 of the error diffusion processor 52 performs Since V1, V2, and V3 are all equal, the input image data is binarized to either "0" or "3". On the other hand, for the pixel of interest whose matrix value of the variation threshold matrix corresponds to “1”, the variation thresholds V1, V2, and V3 have different values. Quaternize to one of the values.
[0081]
As understood from FIG. 15 in which attention is paid to the range of image data values 0 to 15, in this embodiment, the variation thresholds V1, V2, and V3 are varied according to the value (density value) of the image data of the pixel of interest. To be set. Specifically, when the value of the image data is in the range of 0 to 7, the variation thresholds V1, V2, and V3 are intermediate values of the gradations of 0 to 255 regardless of the matrix values of the variation threshold matrix. 127 ". As a result, a normal type error diffusion process in which the quantization threshold does not fluctuate is performed, and the quantization processing unit 63 of the error diffusion processing unit 52 binarizes the input image data into “0” or “3”. I do.
[0082]
When the value of the image data of the target pixel is in the range of 8 to 15, the variation thresholds V1, V2, and V3 are such that the variation width corresponding to the matrix value of the variation threshold matrix is 16 or more. It is set to be smaller than the case.
As described above, in the extremely low density region, the dot concentration can be reduced or eliminated by reducing or eliminating the change in the threshold value according to the matrix value of the variation threshold value matrix. As a result, in the low-density region in the photographic image, large dots are not conspicuous in the output image, and a beautiful low-density gradation image can be reproduced. At the same time, it is possible to suppress the occurrence of a texture due to a change in the quantization threshold.
[0083]
The variation threshold selection unit 84 sets the variation thresholds V1, V2, and V3 based on the image data of the pixel of interest and the matrix value (variation threshold selection information) corresponding to the position of the pixel of interest in the variation threshold matrix. The data is read out from the memories 81, 82, and 83 and supplied to the threshold value selection unit 73.
When the value of the region separation signal is “0” and the fixed thresholds F1, F2, and F3 set by the fixed threshold setting unit 71 are used, the correction value calculation unit 74 sets the generated error correction value EC to “ Set to "0". On the other hand, when the segmentation signal is a value “1” indicating that it is a photograph region, the correction value calculation unit 74 obtains the variation threshold values V1, V2, and V3 from the threshold value selection unit 73, and Then, the generated error correction value EC is calculated.
[0084]
Generation error correction value = (V1 + V2 + V3) ÷ 3-127 (13)
That is, the intermediate gradation value “127” of the average value of the variation threshold values V1, V2, and V3(Intermediate gradation value that divides the entire gradation range into two)Is obtained as an error correction value EC, which is provided to the error correction unit 65. The error corrector 65 calculates the corrected error by subtracting the generated error correction value EC from the error calculated by the error calculator 64 (see the above equation (11)).
[0085]
In this way, when the fluctuation threshold values V1, V2, and V3 that periodically fluctuate in the two-dimensional direction on the image plane are used, the generated error distributed to the surrounding pixels may be affected by the fluctuation of the quantization threshold. Is corrected so that there is no This correction process is, after all, a process of changing the calculation reference value of the generated error according to the variation of the quantization threshold. That is, the minimum gradation value “0”, the first intermediate gradation value “85”, the second intermediate gradation value “170”, and the maximum gradation value “255” in the above equation (10) are based on the calculation criteria of the occurrence error. Value. In the processing of the formula (11) for correcting the generated error, the calculation reference values “0”, “85”, “170”, and “255” of the generated error are corrected by the generated error correction value EC. This is nothing but a process of calculating a quantization error based on the corrected reference values “0-EC”, “85-EC”, “170-EC”, and “255-EC”. As can be understood from the above equation (13), the generated error correction value EC periodically fluctuates along with the periodic fluctuation of the fluctuation thresholds V1, V2, V3. The period fluctuates in the same phase as V3. As a result, it is possible to prevent a large generated error from being distributed to peripheral pixels, and it is possible to efficiently concentrate dots and form good halftone dots.
[0086]
FIG. 16 is a diagram showing the periodic fluctuation of the quantization threshold value and the periodic fluctuation of the error calculation reference value corresponding thereto. When the quantization thresholds (fluctuation thresholds) V1, V2, and V3 periodically fluctuate as shown by the thick lines in FIG. 16, the error correction value EC, which is the deviation of the quantization threshold from the intermediate gradation value "127", is calculated as shown in FIG. 16 as indicated by the arrow. In this case, a downward arrow indicates that the generated error correction value EC is a negative value, and an upward arrow indicates that the generated error correction value EC is a positive value.
[0087]
For example, it is assumed that the variation threshold V1 = V2 = V3 corresponding to the variation threshold matrix value “0” is set smaller than the intermediate value “127”. In this case, the target pixel X having a value smaller than the intermediate value “127” is quantized to a value “3” corresponding to a black pixel when the value is larger than the variation thresholds V1, V2, and V3. . At this time, the error calculation unit 64 calculates a value represented by the following equation as a generated error before correction.
Generation error before correction = value after addition of cumulative error of target pixel X−255 (14)
Therefore, in this case, an occurrence error of an absolute value larger than “127” occurs. The generated error before correction is corrected to a small absolute value by subtracting the generated error correction value EC. As a result of this correction, the generated error after correction is corrected to a small value according to the quantization thresholds V1, V2, V3. This correction is equivalent to correcting the calculation reference value of the occurrence error to a value R (3) smaller than the maximum gradation value “255”.
[0088]
In FIG. 16, the actual error calculation reference values when the image data of the target pixel are quantized into output data “0”, “1”, “2”, and “3” are R (0), R (1), R (2), and R (3). When the variation threshold matrix value is "1", the quaternization process using the variation thresholds V1, V2, and V3 of different values is performed, and accordingly, four error calculation reference values R (0), There are R (1), R (2), and R (3).
[0089]
FIG. 17 shows the effect of quantizing at a higher quantization level only for a particular value (“1” in this embodiment) in the variation threshold matrix (ie, quaternizing in this embodiment). FIG. As described above, by applying the variation threshold matrix shown in FIG. 14A, the quantization threshold can be periodically varied in the two-dimensional direction, and as a result, a halftone dot having a fixed number of lines (for example, 141 lines) can be obtained. The gradation of a photographic image can be expressed. In this case, when a dot DC that is a nucleus of a halftone dot is formed, if the pixels around the DC of the dot are quaternized, the way in which the DC of the dot forming the halftone dot expands or shrinks becomes It can be set more minutely than when binarizing pixels around DC. That is, for example, in the output unit 4 including the laser scanning unit, the laser emission time or the laser emission intensity per pixel is controlled in four stages based on the image data expressed in four values (2 bits). The size of the storage pixel in the section 4 can be controlled in four stages.
[0090]
In this embodiment, pixels near the saddle part (corresponding to “0” in the variation threshold matrix) in the periodic variation of the quantization threshold (corresponding to the matrix value “1” in the variation threshold matrix) are quaternized. Processing is to be performed. As a result, the number of stages of how the dots expand or contract can be determined in detail as compared with the case where the quantization level is fixed to only binarization, and a smooth gradation expression can be achieved. .
In this embodiment, the quaternization process is performed on the pixel “1”, but this is only an example. For example, the quantization of a high quantization level (the In the embodiment, a quaternization process may be performed, or a quantization process with a high quantization level may be performed for both the matrix values “1” and “2”.
[0091]
As described above, according to this embodiment, the total TOTAL_SUM of the differences in the image data between the pixels in the first image block (the matrix of 13 × 7 pixels) including the target pixel, and the image data in the first image block Can be separated and extracted from the pixels constituting the other area based on the magnitude comparison result with the threshold value TH (AV) determined according to the average value AV.
Further, the total sum H_SUM of image data differences in the main scanning direction and the total sum V_SUM of image data differences in the sub-scanning direction of the pixels in the second image block (matrix of 5 × 5 pixels) are determined by predetermined thresholds HL, HH, VL. , VH, it is possible to prevent the pixels constituting the thin line region from being erroneously determined to belong to the photographic region. Therefore, pixels belonging to a photographic image can be extracted very accurately.
[0092]
Further, by performing the above-described area separation processing on the image data that has been subjected to the smoothing processing by the pre-separation filter processing unit 39, the halftone dot area having a high number of lines (175 lines or more) is converted into a photographic area. Can reliably be determined. Accordingly, for a halftone image having a high screen ruling that does not cause moire due to the halftoning processing (improved error diffusion processing) in the halftone processing section 37, an image having good gradation characteristics like a photographic image Playback can be performed.
[0093]
Furthermore, in this embodiment, the pixels determined to belong to the photographic area are subjected to the improved error diffusion processing using the fluctuation threshold values V1, V2, and V3 that periodically fluctuate in the two-dimensional direction. The halftone processing can be applied to a halftone image and a photographic image having a high screen ruling of 175 lines or more. As a result, it is possible to perform dot concentration type halftone processing which is advantageous in image formation by an electrophotographic process, and it is possible to obtain a reproduced image excellent in gradation expression.
[0094]
On the other hand, the normal error diffusion processing in which the quantization threshold is fixed to a constant value is performed on the pixels determined not to belong to the photograph area, so that the character image, the line drawing image, the halftone image (low line) With respect to (number), images can be reproduced with good resolution, and moire fringes do not occur.
Further, in the improved error diffusion processing of this embodiment, since the calculation reference value of the generated error is changed in the same phase as the change of the quantization threshold, a large quantization error due to the change of the quantization threshold is noticed. It is not distributed to peripheral pixels around the pixel. Therefore, the black pixels that should form the halftone dots can be effectively concentrated, so that the halftone processing by the improved error diffusion processing can be favorably performed. In addition, there is an advantage that the density of the image can be reliably stored.
[0095]
Furthermore, in this embodiment, the number of steps of how to expand or contract a halftone dot is increased by performing quaternization processing instead of binarization processing on pixels around a dot serving as a halftone dot. Therefore, it is possible to express a halftone dot image and a photographic image having a high screen ruling with smooth gradation.
As described above, one embodiment of the present invention has been described, but the present invention can be embodied in other forms. For example, in the above-described embodiment, although the threshold matrix value to which a high quantization level (quaternization) is applied is fixed, the third image block including the target pixel (for example, the same as the variation threshold matrix). The average value of the image data in the third image block is calculated for a matrix area having a size, and a matrix value to be subjected to a quantization process (a quaternization process) at a high quantization level based on the average value May be variably set. In this case, the relationship between the average value of the image data in the third image block and the matrix value may be set as shown in Table 1 below when the variation threshold matrix of FIG. Good.
[0096]
[Table 1]

[0097]
In this manner, in any density region, the pixels constituting the periphery of the dot (dot) can be surely subjected to the quaternary processing, so that better tone reproducibility can be obtained. Can be realized.
In the above-described embodiment, the generation error calculation reference value that periodically fluctuates with the fluctuation of the quantization threshold has a constant amplitude with respect to the value of the fluctuation threshold. May vary depending on the size of the image data. For example, by increasing the amplitude in the low-density region, the concentration of dots can be suppressed, and a sense of discomfort due to the concentration of dots can be suppressed, and a beautiful reproduced image can be realized. On the other hand, in the high density region, by setting the amplitude to be small, dots can be satisfactorily concentrated, and good tone reproducibility can be realized.
[0098]
Further, in the above-described embodiment, the quantization level is switched between two levels, that is, binarization and quaternization, but the quantization level is switched between three levels.Less thanMay be done on top. For example, the pixels around the pixel at which the dot nucleus is formed by the binarization processing may be binarized, and the pixels around the pixel may be binarized.
In the above-described embodiment, four-valued pixels around black pixels (nuclei of dots forming halftone dots) are used.ConversionReasonToAlthough this is performed, a pixel to be quaternized may be arranged around a white pixel, and a gray pixel (a small area black pixel) may be formed around the white pixel. That is, in this case, quantization processing at a high quantization level is performed on pixels near the vertex of the fluctuation cycle of the quantization threshold.
[0099]
Further, in the above embodiment, an example in which the present invention is applied to a digital copying machine has been described. However, the present invention can be widely applied to an apparatus for processing image data, for example, a facsimile apparatus and an input apparatus. The present invention can also be applied to a printer that performs image processing on image data to form an image on a recording sheet.
In addition, various design changes can be made within the scope of the matters described in the claims.
[Brief description of the drawings]
FIG. 1 is a block diagram illustrating an electrical configuration of a digital copying machine as an example of an image forming apparatus to which an image processing apparatus according to an embodiment of the present invention has been applied.
FIG. 2 is a block diagram illustrating a functional configuration of an image processing unit.
FIG. 3 is a flowchart for explaining processing contents of an area separation function unit.
FIG. 4 is a diagram illustrating an example of a first image block (a matrix of 13 × 7 pixels) used for extracting pixels in a photograph area.
FIG. 5 is a diagram showing an example of a second image block (5 × 5 pixel matrix) used for extracting pixels in a thin line region.
FIG. 6 is a graph for explaining a setting example of a threshold value TH (AV) for determining a photograph area corresponding to an average value AV of image data of pixels in a first image block.
FIG. 7 is a diagram illustrating a distribution example of a relationship between a sum of density differences H_SUM in a main scanning direction and a sum of density differences V_SUM in a sub-scanning direction in a second image block.
FIG. 8 is a diagram illustrating a configuration of a pre-separation filter processing unit.
FIG. 9 is a diagram for explaining processing performed by a pre-separation filter processing unit;
FIG. 10 is a diagram illustrating the size of a matrix of a first image block.
FIG. 11 is a block diagram illustrating a functional configuration of a halftone processing unit.
FIG. 12 is a diagram for explaining accumulation and distribution of quantization errors.
FIG. 13 is a block diagram illustrating a functional configuration of a parameter generation processing unit.
FIG. 14 is a diagram showing a variation threshold matrix stored in a variation threshold matrix memory and addresses of the variation threshold matrix memory corresponding to each matrix position of the variation threshold matrix.
FIG. 15 is a diagram showing a setting example of fluctuation threshold tables T1, T2, and T3.
FIG. 16 is a diagram showing a periodic variation of a quantization threshold and a periodic variation of an error calculation reference value corresponding thereto.
FIG. 17 is a diagram for explaining the effect of quantizing a specific value in a fluctuation threshold matrix at a higher quantization level.
FIG. 18 is a diagram illustrating a change in an error (quantization error) caused by a change in a quantization threshold.
[Explanation of symbols]
1 Reading unit
3 Image processing unit
4 Output section
37 Halftone processing section
38 Area separation function part
39 Pre-separation filter processing unit
40 Judgment threshold table
51 Parameter Generation Processing Unit
52 error diffusion processing unit
61 Error addition processing unit
62 Accumulated error memory
63 Quantization processing unit
64 error calculator
65 Error correction unit
66 Error distribution processing unit
71 Fixed threshold setting unit
72 Fluctuation threshold setting unit
73 Threshold selector
74 Correction value calculator
80 Fluctuation threshold matrix memory
81 Fluctuation threshold memory
82 Fluctuation threshold memory
83 Fluctuation threshold memory
84 Fluctuation threshold selector
85 Fluctuation threshold selection information generator
86 Fluctuation threshold generator
T1 fluctuation threshold table
T2 fluctuation threshold table
T3 fluctuation threshold table

Claims (5)

An image processing apparatus that quantizes image data representing the density gradation of pixels forming an image into binary or more discrete value data smaller than the maximum density gradation value,
Quantization means for quantizing the image data of the target pixel using a quantization threshold,
Threshold setting means for setting the quantization threshold in the quantization means by changing the cycle periodically;
Correction value calculation means for calculating, as an occurrence error correction value, a deviation of the quantization threshold value set by the threshold value setting means from an intermediate gradation value that divides the entire gradation range into two,
From the image data by subtracting the computed generated error correction value by the tone value and the correction value calculating means corresponding to the discrete value data after quantization, quantization that occurs during the quantization by the quantization means Generating error calculating means for calculating a generating error, which is an error ,
Characterized in that the generating error calculated by the generating error calculating means, in the neighborhood of the pixel of interest is distributed to surrounding pixels which have not yet been quantized, and a error diffusion means for changing the image data of the surrounding pixels Image processing apparatus.   The threshold setting means sets a plurality of quantization thresholds,
  2. The apparatus according to claim 1, wherein the correction value calculating means calculates a deviation of an average value of the plurality of quantization threshold values set by the threshold value setting means from the intermediate gradation value as an occurrence error correction value. An image processing apparatus as described in the above. The threshold value setting means switches between a fixed value and a variable value of the quantization threshold value in the quantization means in accordance with an area separation signal indicating which type of image area the target pixel belongs to. the image processing apparatus according to claim 1 or 2, characterized in that. An image processing method for quantizing image data representing density gradations of pixels constituting an image into binary or more discrete value data smaller than a maximum density gradation value,
A quantization step of quantizing the image data of the pixel of interest using a quantization threshold,
A threshold setting step of setting the quantization threshold by changing the cycle periodically;
A correction value calculation step of calculating, as an occurrence error correction value, a deviation of the quantization threshold value set in the threshold value setting step from an intermediate gradation value that divides the entire gradation range into two,
From the image data, the gradation value corresponding to the discrete value data after quantization and by reducing the correction value generating error correction value calculated by the calculation step, a quantization generated during the quantization by the quantization step A generation error calculation step of calculating a generation error that is an error ,
An image processing method of the calculated generation error, in the neighborhood of the pixel of interest is distributed to surrounding pixels which have not yet been quantized, characterized in that it comprises a error diffusion step of changing the image data of the surrounding pixels . The threshold setting step switches between fixing the quantization threshold in the quantization step and setting the quantization threshold by changing the cycle in accordance with an area separation signal indicating which type of image area the target pixel belongs to. 5. The image processing method according to claim 4 , comprising a step.

Images