JP3741982B2

JP3741982B2 - Noise reduction circuit

Info

Publication number: JP3741982B2
Application number: JP2001274151A
Authority: JP
Inventors: 一憲渡邉
Original assignee: Sharp Corp
Current assignee: Sharp Corp
Priority date: 2001-09-10
Filing date: 2001-09-10
Publication date: 2006-02-01
Anticipated expiration: 2021-09-10
Also published as: JP2003085553A

Description

【０００１】
【発明の属する技術分野】
本発明は、画像又は、映像に混在する雑音成分を除去するノイズリダクション回路に関する。
【０００２】
【従来の技術】
従来のノイズリダクション回路としては、特開平５−１０３２２６号公報や特開平８−２２５３９号公報に記載されたものがある。
【０００３】
特開平５−１０３２２６号公報に記載のノイズリダクション回路では、まず、横、縦、斜め各方向の画素での平滑化信号と相関性を平滑化回路、相関検出回路においてそれぞれ検出する。そして、検出された各方向での相関検出信号は、輪郭方向判定回路において比較され、最も相関検出信号が小さい方向に沿って輪郭があるものとし、その相関検出信号の小さな方向と同じ方向の画素を用いた平滑化信号を４入力セレクタで選択し、出力するものが記載されている。
【０００４】
特開平８−２２５３９号公報に記載のノイズリダクション回路では、まず、平滑化マトリクスＦｉ（ｉ＝１、２…）の全てを含むことができる最小サイズのマトリクスをＭとする。ここで、マトリクスＭの各成分は、各平滑化マトリクスＦｉ（ｉ＝１、２…）の対応する成分から絶対値が最大なものを選ぶ。このマトリクスＭによる線形フィルタの周波数特性から、ゼロクロスする周波数すべてを含む帯域Ｉを決定する。線形Ｌｏｗ−ｐａｓｓフィルタＬとして領域Ｉを含む高周波側を減衰させるものと、線形高域強調フィルタＨとしてフィルタＬとフィルタＨによる合成フィルタＬＨがちょうど領域Ｉを強調するものを決定する。そして、画像データに合成フィルタ処理ＬＦＨを施すことによってノイズ低減を行うものが記載されている。
【０００５】
【発明が解決しようとする課題】
しかしながら、従来のノイズリダクション回路は、エッジがない部分や複数のエッジが重なっている部分でのノイズの低減効果が低い場合があった。
【０００６】
本発明は、前記の問題点を解消するためになされたものであって、エッジがない部分や複数のエッジが重なっている部分においてもノイズの低減効果を高めることのできるノイズリダクション回路を提供することを目的とする。
【０００７】
【課題を解決するための手段】
本発明は、上記の目的を達成するため、次の構成を有する。
本発明の第１の要旨は、所定大きさの処理領域を、注目画素と該注目画素の近傍画素を含む複数の小領域に分け、各小領域毎に所定画素値に関する平均値である小領域画素平均値を出力する平均値出力手段と、前記平均値出力手段の各小領域毎に、前記小領域画素平均値と各画素の所定画素値との残差を求める残差処理と、該残差処理後の処理結果の平均値を求める残差平均値処理より、各小領域毎の画素値の分散を示す第１の判定値を出力する第１の判定手段と、前記第１の判定値の内、最も小さな処理結果を示す小領域に対応する前記小領域画素平均値を基準値として出力する基準値出力手段と、前記基準値と前記小領域画素平均値との残差を求める残差処理と、該残差処理後の処理結果の平均値を求める残差平均値処理より前記基準値に対する各小領域の分散を示す第２の判定値を出力する第２の判定手段と、前記小領域画素平均値と、前記第１の判定値と、前記第２の判定値とを用いて、前記注目画素の所定画素値に関する評価値を出力する評価値出力手段とを有するノイズリダクション回路にある。
【０００８】
本発明の第２の要旨は、前記評価値出力手段は、前記第１の判定値が第１の閾値より小さく、且つ、前記第２の判定値が第２の閾値より小さい該当小領域の小領域画素平均値の平均値を、前記評価値として出力することを特徴とする要旨１記載のノイズリダクション回路にある。
【０００９】
本発明の第３の要旨は、前記評価値出力手段は、前記第１の判定値が第１の閾値より小さく、且つ、前記第２の判定値が第２の閾値より小さい該当小領域の小領域画素平均値を求め、前記該当小領域の小領域画素平均値から前記基準値となっている少領域の小領域画素平均値を除いた小領域画素平均値と、前記基準値との平均値を前記評価値として出力することを特徴とする要旨１記載のノイズリダクション回路にある。
【００１０】
本発明の第４の要旨は、前記評価値出力手段は、前記第１の判定値が第１の閾値より小さい該当小領域の小領域画素平均値と前記基準値の差分値を求める処理と、第２の閾値に対して、前記該当小領域に対応する前記第２の判定値の重み付け係数を求める処理と、前記差分値と前記重み付け係数とを、小領域毎に積を求める重み付け処理とを行い、前記基準値に前記重み付け処理の結果を加えることで前記評価値を求めることを特徴とする要旨１記載のノイズリダクション回路にある。
【００１１】
本発明の第５の要旨は、前記第１の判定手段又は第２の判定手段は、前記残差処理と前記残差平均値処理により分散処理、標準偏差処理、或は差分絶対値平均処理を行うことを特徴とする要旨１乃至３記載のノイズリダクション回路にある。
【００１２】
本発明の第１の要旨によれば、第１の判定値により各小領域が平坦なのかエッジが存在するのかを判定できる。更に、複数の平坦な部分が存在しても最も平坦な部分の小領域の平均値を基準値として、再度、第２の判定値により各小領域が平坦なのかエッジが存在するのかを判定できる。そして、前記平均値と前記第１の判定値と前記第２の判定値とを用いて評価値出力手段は、複数のエッジの存在を判定できる。
通常、ノイズはランダムであり、平均値をとることによりノイズの低減効果があり、第１の判定手段、第２の判定手段により平均値を取ることでノイズの低減効果を大きくできる。
【００１３】
本発明の第２、第３の要旨によれば、評価値出力手段は、第１の判定値が第１の閾値より小さく、且つ、第２の判定値が第２の閾値より小さい該当小領域の小領域画素平均値の平均値を、評価値として出力することで、エッジが少ない部分や複数のエッジが重なっている部分においてもノイズの低減効果を上げることができる。
【００１４】
本発明の第４の要旨によれば、評価値出力手段は、第１の判定値が第１の閾値より小さい該当小領域の小領域画素平均値と基準値の差分値を求める処理と、第２の閾値に対して、該当小領域に対応する第２の判定値の重み付け係数を求める処理と、差分値と重み付け係数とを、小領域毎に積を求める重み付け処理とを行い、基準値に重み付け処理の結果を加えることで評価値を求めることで、エッジが少ない部分や複数のエッジが重なっている部分においてもノイズの低減効果を上げ、且つ、鮮鋭感の低下が少ない効果がある。
【００１５】
本発明の第５の要旨によれば、前記第１の判定手段又は第２の判定手段は、前記残差処理と前記残差平均値処理により分散処理、標準偏差処理、或は差分絶対値平均処理を行うことで、第１の判定手段における各小領域毎の画素値の分散又は、第２の判定手段における基準値に対する各小領域の分散を確実に求めることができる。
【００１６】
【発明の実施の形態】
以下、図面を参照して本発明の実施形態を詳細に説明する。
図１は、本発明の実施の形態に係るノイズリダクション回路１００のブロック図である。ノイズリダクション回路１００は、信号入力部１と、ラインメモリ２と、平均値出力回路３と、第１の判定回路４と、基準値出力回路５と、第２の判定回路６と、評価値出力回路７と、信号出力部８とを有し、映像信号又は画像信号に含まれる雑音を低減する。
【００１７】
ラインメモリ２は、信号入力部１から供給される映像信号又は、画像信号を水平走査期間分だけ記憶し、遅らせるためのメモリである。
【００１８】
平均値出力回路３は、ラインメモリ２からの映像信号又は画像信号を、注目画素とその近傍画素を注目画素を含んだ複数の小領域に分け、各小領域での画素の特徴量である画素値（例えば、画素の濃度、明度、輝度、色差、或は強度等）の平均値を出力する。強度、濃度、輝度、色差、及び明度の規定は、既知の量子化における諧調表現に使用できるものであればよく、例えば、強度は光の強さ又は反射率により求められ、濃度は強度の対数を取り、符号を反転させることで求められ、明度はＬａｂやＬｕｖの均等色空間で使用される明度を用い、輝度や色差は映像信号で用いられる輝度や色差を用いることができる。
【００１９】
第１の判定回路４は、平均値出力回路３からの各領域毎の画素値の平均値と、ラインメモリ２からの領域内の各画素の画素値とから、残差二乗和を用いる各領域の分散又は標準偏差等を出力する。
【００２０】
基準値出力回路５は、第１の判定回路４からの各領域の判定出力Ａと、平均値出力回路３からの各領域の平均値とから、最も平坦な領域の平均値を基準値として出力する。
【００２１】
第２の判定回路６は、基準値出力回路５からの基準値と、平均値出力回路３からの各領域の平均値とから、残差二乗和を用いる各領域の分散又は標準偏差等を出力する。
【００２２】
評価値出力回路７は、平均値出力回路３からの各領域の平均値と、第１の判定回路４からの各領域の判定出力Ａと、基準値出力回路５からの基準値と、第２の判定回路６からの判定出力Ｂとを用いて評価値を出力する。
【００２３】
映像信号や画像信号における信号対ノイズ比（Ｓ／Ｎ）は、下記式（１）に示すように信号Ｓは実値であり、ノイズＮは実効値であるため、サンプル数が多くなればなるほどＳ／Ｎ比は大きくなる。注目画素とその近傍画素ｎ個での平均値は下記式（２）で表され、平均値を取る画素の個数が多くなる程、その時のＳ／Ｎが良くなる。
【００２４】
【数１】

【００２５】
本実施の形態では、注目画素とその近傍画素を複数の小領域に分け、平均値出力回路３により平均値を、第１の判定回路により各小領域の平坦度を検出する。ノイズの低減効果が最も出ている小領域は、最も平坦な小領域となるので、基準値出力回路５でその小領域の平均値を基準値として選択する。
さらにノイズの低減効果を得る為に、平均値を取る画素数を多くするため、第２の判定回路６により基準値に近い平均値を出している少領域を判定し、評価値出力回路７で評価値として出力する。各回路の動作により、平坦でない小領域（エッジの在る部分）や、平坦でも注目画素の値から大きく離れた平均値を出す小領域を除いた領域での平均値を評価値として出力する。場合によっては全ての小領域が評価値を出力するのに使用される。
【００２６】
次に、図２〜図５を参照して、図１に示したノイズリダクション回路１００の動作をより詳細に説明する。
【００２７】
図２は、注目画素とその近傍画素のｎ×ｎ画素の範囲（ｎは３以上の奇数）における、注目画素を含む小領域Ａ１〜Ａ９の概略位置を示している。尚、小領域の大きさや数は、ｎの数により異なる。
本実施形態での小領域Ａ１〜Ａ９は、注目画素から紙面上上方への小領域Ａ１と、注目画素から紙面上右斜め上方への領域Ａ２、注目画素から紙面上右方向への領域Ａ３、注目画素から紙面上右斜め下方への領域Ａ４、注目画素から紙面下方への領域Ａ５、注目画素から紙面上左斜め下方への領域Ａ６、注目画素から紙面上左方向への領域Ａ７、注目画素から紙面上右上斜め上方へのＡ８、及び注目画素とその周辺の領域Ａ９としている。
【００２８】
図３は、注目画素とその近傍画素の５×５画素の範囲を小領域Ａ１〜Ａ９に分けた場合を示している。図３では、各小領域Ａ１〜Ａ９を示し位置に「１」を、それ以外を「０」と示している。
【００２９】
小領域Ａ１は、５×５画素の中心位置にある注目画素Ｔから紙面上上方への３画素及び該画素に面する４画素からなっている。
【００３０】
同様に、領域Ａ２は注目画素ＴＧから紙面右斜め上方への３画素及び該画素に面する４画素からなり、領域Ａ３は注目画素ＴＧから紙面右方向への３画素及び該画素に面する４画素からなり、領域Ａ４は注目画素ＴＧから紙面右斜め下方への３画素及び該画素に面する４画素からなり、領域Ａ５は注目画素ＴＧから紙面下方への３画素及び該画素に面する４画素からなり、領域Ａ６は注目画素ＴＧから紙面左斜め下方への３画素及び該画素に面する４画素からなり、領域Ａ７は注目画素ＴＧから紙面左方向への３画素及び該画素に面する４画素からなり、領域Ａ８は注目画素から紙面右上斜め上方への３画素及び該画素に面する４画素からなり、及び領域Ａ９は注目画素ＴＧと該画素ＴＧに面する８画素からなっている。
【００３１】
図４は、ラインメモリ２から出力される注目画素とその近傍画素の５×５画素の範囲内の濃度の一例を示しており、係る画像に基づくノイズリダクション回路１００内部でのデータ処理結果を図５に示す。
【００３２】
図５は、最上段に前記図２，３で示す領域番号を、２段目に各領域毎の平均値出力回路３から出力される各平均値を、３段目に第１の判定回路４から出力される各領域毎の判定値Ａを、４段目に基準値出力回路５から出力される基準値を、５段目に第２の判定回路６から出力される各領域毎の判定値Ｂを、及び最下段に評価値出力回路７から出力される評価値を示している。
【００３３】
前記平均値出力回路３から出力される平均値ｅは、図４に示す画素濃度を示す場合に、図３に示す各領域Ａ１〜Ａ９毎の平均濃度を求めるものであり下記数式（３）により求められる。例えば、領域Ａ１の場合には「１０、１０、５０、５０、５０、５０、５０」の７画素の平均濃度であり四捨五入して「ｅ＝３９」と求められる。
尚、平均値出力回路３の平均値出力は、一実施形態であり他の方法、例えばメディアンフィルタでも良い。
【００３４】
【数２】

【００３５】
前記第１の判定回路４から出力される各領域毎の判定値Ａは、平均値出力回路３からの平均値、例えば領域Ａ１の平均値「ｅ＝３９」と、領域Ａ１内の各画素の濃度「Ｐ＝１０、１０、５０、５０、５０、５０、５０」とから、残差二乗和を用いる下記数式（４）により分散σを求め、四捨五入して「３２７」と求められる。
尚、判定値Ａとして標準偏差ｄを求める場合には、下記数式（５）に示すように分散σのルートとなる。
【００３６】
【数３】

【００３７】
また、判定回路４の判定値Ａは、他の方法、例えば平均値ｅと領域内の各画素Ｐｎとの差（残差）の絶対値の平均（差分絶対値平均処理）でも良い。
【００３８】
基準値出力回路５から出力される基準値は、判定値Ａが最も小くなる、図５の場合「０」を示す、最も平坦な領域Ａ２の平均値ｅ＝５０となる。
【００３９】
第２の判定回路６から出力される各領域毎の判定値Ｂは、基準値出力回路５からの基準値「５０」と平均値出力回路３からの各領域Ａ１〜Ａ９の平均値ｅとの分散σを示す。例えば、領域Ａ１の場合には、上記式（４）より、判定値Ｂ＝（５０−３９）²＝１２１となる。
尚、判定値Ｂは、標準偏差や基準値と各平均値との差の絶対値でも良い。
【００４０】
前記評価値出力回路７から出力される注目画素Ｔの評価値は、第１の判定回路４からの判定値Ａの値がある閾値（本実施形態では「２０」）以下の領域で、かつ判定回路６からの判定値Ｂの値がある閾値（本実施形態では「２０」）以下の領域、すなわち領域Ａ２、Ａ３の平均値出力回路３の平均値「５０」と「４７」と、基準値出力回路５からの基準値「５０」とから、基準値「５０」となっている領域Ａ２を除いて、平均値（この場合、基準値「５０」と領域３の平均値「４７」との計算となるので、（５０＋４７）／２＝４９）を出力する。
【００４１】
通常ノイズはランダムであり、平均値をとることによりノイズの低減効果があることが知られており、平均値を取る数が多ければ多いほど効果が大きくなる。
本実施の形態では、評価値出力回路７は各領域の最も平坦な領域を基準値とし、その基準値に近く、かつ平坦な部分の領域での平均値をとる事により、ノイズの低減効果を上げるように働くこととなる。
【００４２】
すなわち、評価値出力回路７は、図２のような複数のエッジを含んだ部分（濃度が５０、４０、１０の境界を含む領域）でも、第１の判定回路４からの各領域の判定出力Ａと第２の判定回路６からの各領域の判定出力Ｂによりエッジの領域を除いた平坦な領域の平均値となるためノイズの低減効果が得られる。
【００４３】
次に、第２の実施の形態に係るノイズリダクション回路について図６，７を参照しつつ説明する。尚、本第２の実施形態は、上記第１の実施形態とは評価値出力回路での評価値の演算処理内容を相違するものであり、以下、該相違部分を中心に説明し、前記した構成と同一部分には同一符号を示す。
【００４４】
本実施形態の評価値出力回路７では、まず、第１の判定回路４からの値がある閾値、本実施形態では「２０」以下の領域Ａ２，Ａ３，Ａ４の平均値ｅ（５０，４７，４３）と基準値「５０」との差分値Ｄ（０，−３，−７）を求める。
【００４５】
次に、上記選択された領域Ａ２，Ａ３，Ａ４の第２の判定回路６からの判定値Ｂに重み付け係数ｙを掛ける。重み付け係数は、例えば図７に示すように、判定値Ｂを横軸、重み係数ｙを縦軸とし、判定値Ｂの閾値を閾値ｃとした場合に、下記重み関数ｙとする。
ｙ＝−Ｂ／ｃ＋１（判定値Ｂが閾値ｃ以下の場合）
ｙ＝０（判定値Ｂが閾値ｃより大きい場合）
本実施形態では、閾値ｃ＝「１０」とした場合、領域Ａ２，Ａ３，Ａ４の判定値Ｂ（０、９、４９）の重み係数ｙは各々（１、０．１、０）となる。
【００４６】
次に、上記差分値Ｄ（０，−３，−７）と重み係数ｙ（１、０．１、０）を各領域Ａ２，Ａ３，Ａ４毎に掛け、重み付け差分値ＷＤ（０×１，−３×０．１、−７×０）を求める。
【００４７】
次に、基準値となっている領域Ａ２の重み付け差分値を除いて、基準値５０に重み付け差分値ＷＤを加え、（５０＋（−０．３）＋０＝４９．７）で四捨五入して「５０」を評価値として出力する。
【００４８】
本第２実施形態と前記第１の実施形態とは、平均値をとることによるノイズの低減効果は変わらないが、平均値をとる数が多ければ多いほど鮮鋭感が無くなる事があるので、この回路はある程度のノイズの低減効果がありかつ鮮鋭感の低下が少ない動作を実現できる。
【００４９】
なお、前記の実施形態では本発明の好適例を説明したが、本発明はこれに限定されないことはもちろんである。
例えば、上記実施の形態では、信号入力部１から供給される映像信号又は、画像信号を水平走査期間分だけ記憶し、遅らせるためにラインメモリ２を用いたが、図８に示すようにフィールドメモリ２Ａを用いてもよい。この場合、フィールドメモリ２Ａを利用する事により、映像信号又は画像信号の１画面内で任意に図３の小領域をとることができるので、平均値出力回路３、第１の判定回路４、基準値出力回路５、第２の判定回路６、及び評価値出力回路７を全て、マイコン等によるソフトウエアで行う事もでき、コンパクトな装置とできる。そのため、デジタルカメラ（撮像器機）、携帯情報機器、携帯通信機器等の携帯可能な表示機器等での利用が考えられる。
【００５０】
【発明の効果】
以上説明した通り、本発明の第１の要旨によれば、第１の判定手段により各小領域が平坦なのかエッジが存在するのかを判定し、複数の平坦な部分が存在しても第２の判定手段及び評価値出力手段により複数のエッジの存在を判定できるので、エッジがない部分や複数のエッジが重なっている部分においてもノイズの低減効果を高めることのできる。
【００５１】
本発明の第２、第３の要旨によれば、評価値出力手段により複数のエッジの存在を判定できるので、エッジが少ない部分や複数のエッジが重なっている部分においてもノイズの低減効果がある。
【００５２】
本発明の第４の要旨によれば、エッジが少ない部分や複数のエッジが重なっている部分においてもノイズの低減効果を上げ、且つ、鮮鋭感の低下が少ない効果がある。
【００５３】
本発明の第５の要旨によれば、各小領域の分散を確実に求めることができる。
【図面の簡単な説明】
【図１】本発明の第１の実施形態に係るノイズリダクション回路のブロック図である。
【図２】本発明の第１の実施形態に係る注目画素を含むｎ×ｎ画素範囲における注目画素を含む小領域Ａ１〜Ａ９の概略位置の説明図である。
【図３】本発明の第１の実施形態に係る注目画素を含むｎ×ｎ画素範囲における注目画素を含む小領域Ａ１〜Ａ９の説明図である。
【図４】本発明の第１の実施形態に係るラインメモリ２から出力される注目画素とその近傍画素の５×５画素の範囲内の各画素の濃度の説明図である。
【図５】本発明の第１の実施形態に係る平均値出力回路３、第１の判定回路４、基準値出力回路５、第２の判定回路６、及び評価値出力回路７からの出力信号を示す説明図である。
【図６】本発明の第２の実施形態に係る平均値出力回路３、第１の判定回路４、基準値出力回路５、第２の判定回路６、及び評価値出力回路７からの出力信号を示す説明図である。
【図７】本発明の第２の実施形態に係る評価値出力回路７での重み付け関数の説明図である。
【図８】本発明のその他の実施形態に係るノイズリダクション回路のブロック図である。
【符号の説明】
２ラインメモリ
２Ａフィールドメモリ
３平均値出力回路
４第１の判定回路
５基準値出力回路
６第２の判定回路
７評価値出力回路
１０フィールドメモリ
Ａ１〜Ａ９小領域
ｅ平均値
Ａ，Ｂ判定値
Ｔ注目画素[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a noise reduction circuit that removes noise components mixed in an image or video.
[0002]
[Prior art]
Conventional noise reduction circuits include those described in Japanese Patent Application Laid-Open Nos. 5-103226 and 8-22539.
[0003]
In the noise reduction circuit described in Japanese Patent Laid-Open No. 5-103226, first, smoothing signals and correlations in pixels in horizontal, vertical, and diagonal directions are detected by a smoothing circuit and a correlation detection circuit, respectively. Then, the detected correlation detection signal in each direction is compared in the contour direction determination circuit, and it is assumed that there is a contour along the direction in which the correlation detection signal is the smallest, and pixels in the same direction as the small direction of the correlation detection signal A smoothing signal using the signal is selected by a 4-input selector and output.
[0004]
In the noise reduction circuit described in Japanese Patent Laid-Open No. 8-22539, first, let M be the minimum size matrix that can include all of the smoothing matrices Fi (i = 1, 2,...). Here, for each component of the matrix M, the one having the maximum absolute value is selected from the corresponding components of each smoothing matrix Fi (i = 1, 2,...). From the frequency characteristic of the linear filter based on the matrix M, the band I including all the frequencies that are zero-crossed is determined. The linear low-pass filter L that attenuates the high frequency side including the region I and the linear high-frequency emphasis filter H that the filter L and the synthesis filter LH of the filter H just enhance the region I are determined. And what performs noise reduction by performing synthetic filter processing LFH on image data is described.
[0005]
[Problems to be solved by the invention]
However, the conventional noise reduction circuit sometimes has a low noise reduction effect in a portion where there is no edge or a portion where a plurality of edges overlap.
[0006]
The present invention has been made to solve the above-described problems, and provides a noise reduction circuit capable of enhancing the noise reduction effect even in a portion where there is no edge or a portion where a plurality of edges overlap. For the purpose.
[0007]
[Means for Solving the Problems]
In order to achieve the above object, the present invention has the following configuration.
The first gist of the present invention is to divide a processing area of a predetermined size into a plurality of small areas including a pixel of interest and neighboring pixels of the pixel of interest, and a small area that is an average value for a predetermined pixel value for each small area An average value output means for outputting a pixel average value; a residual process for obtaining a residual between the small area pixel average value and a predetermined pixel value of each pixel for each small area of the average value output means; A first determination means for outputting a first determination value indicating a variance of pixel values for each small region from the residual average value processing for calculating an average value of the processing results after the difference processing; and the first determination value A reference value output means for outputting, as a reference value, the small region pixel average value corresponding to the small region indicating the smallest processing result, and a residual for obtaining a residual between the reference value and the small region pixel average value The reference value is calculated from the residual average value processing for calculating the average value of the processing result and the processing result after the residual processing. Using second determination means for outputting a second determination value indicating the variance of each small region, the small region pixel average value, the first determination value, and the second determination value, The noise reduction circuit includes evaluation value output means for outputting an evaluation value related to a predetermined pixel value of the target pixel.
[0008]
The second gist of the present invention is that the evaluation value output means has a small small area corresponding to the first determination value smaller than the first threshold and the second determination value smaller than the second threshold. The noise reduction circuit according to the first aspect is characterized in that an average value of area pixel average values is output as the evaluation value.
[0009]
The third gist of the present invention is that the evaluation value output means has a small small area corresponding to the first determination value smaller than the first threshold and the second determination value smaller than the second threshold. An area pixel average value is obtained, and an average value of the small area pixel average value obtained by subtracting the small area pixel average value of the small area that is the reference value from the small area pixel average value of the corresponding small area, and the reference value Is output as the evaluation value. A noise reduction circuit according to the first aspect of the invention.
[0010]
According to a fourth aspect of the present invention, the evaluation value output means obtains a difference value between the small region pixel average value of the small region in which the first determination value is smaller than a first threshold and the reference value. A process of obtaining a weighting coefficient of the second determination value corresponding to the corresponding small area with respect to a second threshold, and a weighting process of obtaining a product of the difference value and the weighting coefficient for each small area. The noise reduction circuit according to claim 1, wherein the evaluation value is obtained by performing and adding the result of the weighting process to the reference value.
[0011]
According to a fifth aspect of the present invention, the first determination unit or the second determination unit performs a variance process, a standard deviation process, or a difference absolute value average process by the residual process and the residual average value process. The noise reduction circuit according to the first aspect or the third aspect is characterized in that it is performed.
[0012]
According to the first aspect of the present invention, it is possible to determine whether each small region is flat or has an edge based on the first determination value. Furthermore, even if there are a plurality of flat portions, it is possible to determine again whether each small region is flat or has an edge based on the second determination value using the average value of the small regions of the flattest portion as a reference value. . Then, the evaluation value output means can determine the presence of a plurality of edges using the average value, the first determination value, and the second determination value.
Usually, noise is random, and taking an average value has an effect of reducing noise. Taking an average value by the first determination means and the second determination means can increase the noise reduction effect.
[0013]
According to the second and third aspects of the present invention, the evaluation value output means includes a corresponding small region in which the first determination value is smaller than the first threshold value and the second determination value is smaller than the second threshold value. By outputting the average value of the small region pixel average values as an evaluation value, it is possible to increase the noise reduction effect even in a portion with few edges or a portion where a plurality of edges overlap.
[0014]
According to the fourth aspect of the present invention, the evaluation value output means obtains a difference value between the small region pixel average value and the reference value of the small region in which the first determination value is smaller than the first threshold, For the threshold of 2, a process for obtaining a weighting coefficient of the second determination value corresponding to the corresponding small area and a weighting process for obtaining a product of the difference value and the weighting coefficient for each small area are performed, By obtaining the evaluation value by adding the result of the weighting process, the effect of reducing noise can be improved and the sharpness can be reduced less even in a portion with few edges or a portion where a plurality of edges overlap.
[0015]
According to a fifth aspect of the present invention, the first determination unit or the second determination unit is configured to perform variance processing, standard deviation processing, or difference absolute value average by the residual processing and the residual average value processing. By performing the processing, the variance of the pixel values for each small region in the first determination unit or the variance of each small region with respect to the reference value in the second determination unit can be reliably obtained.
[0016]
DETAILED DESCRIPTION OF THE INVENTION
Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.
FIG. 1 is a block diagram of a noise reduction circuit 100 according to an embodiment of the present invention. The noise reduction circuit 100 includes a signal input unit 1, a line memory 2, an average value output circuit 3, a first determination circuit 4, a reference value output circuit 5, a second determination circuit 6, and an evaluation value output. A circuit 7 and a signal output unit 8 are provided to reduce noise included in the video signal or the image signal.
[0017]
The line memory 2 is a memory for storing and delaying a video signal or an image signal supplied from the signal input unit 1 for a horizontal scanning period.
[0018]
The average value output circuit 3 divides the video signal or image signal from the line memory 2 into a plurality of small regions including the pixel of interest and its neighboring pixels, and is a pixel feature amount in each small region An average value of values (for example, pixel density, brightness, luminance, color difference, or intensity) is output. The definition of intensity, density, brightness, color difference, and lightness may be anything that can be used for gradation expression in known quantization. For example, the intensity is obtained by the intensity or reflectance of light, and the density is a logarithm of intensity. The brightness is obtained by reversing the sign, the brightness is the brightness used in the uniform color space of Lab or Luv, and the brightness and color difference used in the video signal can be used as the brightness and color difference.
[0019]
The first determination circuit 4 uses each of the regions using a residual sum of squares from the average value of the pixel values for each region from the average value output circuit 3 and the pixel value of each pixel in the region from the line memory 2. The variance or standard deviation is output.
[0020]
The reference value output circuit 5 outputs the average value of the flattest area as a reference value from the determination output A of each area from the first determination circuit 4 and the average value of each area from the average value output circuit 3. To do.
[0021]
The second determination circuit 6 outputs the variance or standard deviation of each region using the residual square sum from the reference value from the reference value output circuit 5 and the average value of each region from the average value output circuit 3. To do.
[0022]
The evaluation value output circuit 7 includes an average value of each region from the average value output circuit 3, a determination output A of each region from the first determination circuit 4, a reference value from the reference value output circuit 5, and a second value. The evaluation value is output using the determination output B from the determination circuit 6.
[0023]
The signal-to-noise ratio (S / N) in the video signal or the image signal is such that the signal S is an actual value and the noise N is an effective value as shown in the following formula (1). The S / N ratio increases. The average value of the pixel of interest and n neighboring pixels is expressed by the following formula (2). The greater the number of pixels that take the average value, the better the S / N at that time.
[0024]
[Expression 1]

[0025]
In the present embodiment, the pixel of interest and its neighboring pixels are divided into a plurality of small regions, the average value is detected by the average value output circuit 3, and the flatness of each small region is detected by the first determination circuit. Since the small area where the noise reduction effect is most significant is the flattest small area, the reference value output circuit 5 selects the average value of the small areas as the reference value.
Further, in order to obtain a noise reduction effect, in order to increase the number of pixels taking the average value, the second determination circuit 6 determines a small area that is outputting an average value close to the reference value, and the evaluation value output circuit 7 Output as an evaluation value. By the operation of each circuit, an average value is output as an evaluation value in a small area that is not flat (a portion having an edge) or a small area that is flat but does not have a small area that gives an average value far from the value of the target pixel. In some cases, all subregions are used to output evaluation values.
[0026]
Next, the operation of the noise reduction circuit 100 shown in FIG. 1 will be described in more detail with reference to FIGS.
[0027]
FIG. 2 shows the approximate positions of the small regions A1 to A9 including the target pixel in the range of n × n pixels (n is an odd number of 3 or more) of the target pixel and its neighboring pixels. Note that the size and number of small regions differ depending on the number of n.
The small areas A1 to A9 in the present embodiment include a small area A1 from the target pixel to the upper side of the paper, an area A2 from the target pixel to the upper right side of the paper, an area A3 from the target pixel to the right side of the paper, A region A4 from the target pixel to the lower right side of the paper surface, a region A5 from the target pixel to the lower side of the paper surface, a region A6 from the target pixel to the lower left side of the paper surface, a region A7 from the target pixel to the left of the paper surface A8 from the upper right to the upper right of the drawing, and the target pixel and its surrounding area A9.
[0028]
FIG. 3 shows a case where the 5 × 5 pixel range of the target pixel and its neighboring pixels is divided into small areas A1 to A9. In FIG. 3, the small areas A1 to A9 are shown, and “1” is indicated at the position, and “0” is indicated otherwise.
[0029]
The small area A1 includes three pixels from the target pixel T at the center position of 5 × 5 pixels to the upper side of the paper and four pixels facing the pixel.
[0030]
Similarly, the area A2 is composed of three pixels from the target pixel TG diagonally upward to the right of the paper surface and four pixels facing the pixel, and the area A3 is three pixels from the target pixel TG to the right side of the paper and four pixels facing the pixel. The region A4 is composed of three pixels from the target pixel TG obliquely downward to the right of the paper surface and four pixels facing the pixel, and the region A5 is composed of three pixels downward from the pixel of interest TG and the pixel 4 facing the pixel. The region A6 is composed of three pixels diagonally downward to the left of the paper surface from the target pixel TG and four pixels facing the pixel, and the region A7 is facing three pixels from the target pixel TG to the left side of the paper and the pixel. The area A8 is composed of 3 pixels from the target pixel to the upper right and obliquely upward of the page and 4 pixels facing the pixel, and the area A9 is composed of the target pixel TG and the 8 pixels facing the pixel TG. .
[0031]
FIG. 4 shows an example of the density within the range of 5 × 5 pixels of the target pixel and its neighboring pixels output from the line memory 2, and shows the data processing result inside the noise reduction circuit 100 based on the image. As shown in FIG.
[0032]
FIG. 5 shows the region numbers shown in FIGS. 2 and 3 in the top row, the average values output from the average value output circuit 3 for each region in the second row, and the first determination circuit 4 in the third row. The determination value A for each region output from the reference value, the reference value output from the reference value output circuit 5 at the fourth level, the determination value for each region output from the second determination circuit 6 at the fifth level. B and the evaluation value output from the evaluation value output circuit 7 are shown at the bottom.
[0033]
When the average value e output from the average value output circuit 3 indicates the pixel density shown in FIG. 4, the average density for each of the areas A1 to A9 shown in FIG. 3 is obtained. Desired. For example, in the case of the area A1, the average density of 7 pixels “10, 10, 50, 50, 50, 50, 50” is rounded off and “e = 39” is obtained.
The average value output of the average value output circuit 3 is one embodiment, and other methods such as a median filter may be used.
[0034]
[Expression 2]

[0035]
The determination value A for each region output from the first determination circuit 4 is an average value from the average value output circuit 3, for example, an average value “e = 39” of the region A1, and each pixel in the region A1. From the density “P = 10, 10, 50, 50, 50, 50, 50”, the variance σ is obtained by the following formula (4) using the residual sum of squares, and rounded off to obtain “327”.
When the standard deviation d is obtained as the determination value A, the root of the variance σ is obtained as shown in the following formula (5).
[0036]
[Equation 3]

[0037]
The determination value A of the determination circuit 4 may be another method, for example, an average of absolute values (difference absolute value average processing) of differences (residuals) between the average value e and each pixel Pn in the region.
[0038]
The reference value output from the reference value output circuit 5 is the average value e = 50 of the flattest region A2, which indicates “0” in the case of FIG.
[0039]
The determination value B for each region output from the second determination circuit 6 is the difference between the reference value “50” from the reference value output circuit 5 and the average value e of the regions A1 to A9 from the average value output circuit 3. The variance σ is shown. For example, in the case of the area A1, the determination value B = (50−39) ² = 121 from the above formula (4).
The determination value B may be a standard deviation or an absolute value of a difference between the reference value and each average value.
[0040]
The evaluation value of the target pixel T output from the evaluation value output circuit 7 is an area in which the value of the determination value A from the first determination circuit 4 is equal to or less than a certain threshold value (“20” in the present embodiment) and is determined. The average value “50” and “47” of the average value output circuit 3 in the area where the value of the determination value B from the circuit 6 is equal to or smaller than a certain threshold value (“20” in this embodiment), that is, the reference value The average value (in this case, the reference value “50” and the average value “47” of the region 3) is excluded from the reference value “50” from the output circuit 5 except for the region A2 where the reference value is “50”. Since this is a calculation, (50 + 47) / 2 = 49) is output.
[0041]
Normally, noise is random, and it is known that taking an average value has an effect of reducing noise. The larger the number of average values, the greater the effect.
In the present embodiment, the evaluation value output circuit 7 uses the flattest area of each area as a reference value, and takes an average value in the area of the flat portion that is close to the reference value, thereby reducing the noise. Will work to raise.
[0042]
That is, the evaluation value output circuit 7 outputs the determination output of each region from the first determination circuit 4 even in a portion including a plurality of edges as shown in FIG. 2 (region including the boundaries of

density

50, 40, 10). Since the average value of the flat region excluding the edge region is obtained by A and the determination output B of each region from the second determination circuit 6, a noise reduction effect can be obtained.
[0043]
Next, a noise reduction circuit according to the second embodiment will be described with reference to FIGS. The second embodiment is different from the first embodiment in the evaluation value calculation processing contents in the evaluation value output circuit. Hereinafter, the difference will be mainly described and described above. The same parts as those in the configuration are denoted by the same reference numerals.
[0044]
In the evaluation value output circuit 7 of the present embodiment, first, a value from the first determination circuit 4 is a certain threshold, in this embodiment, an average value e (50, 47, 43) and a difference value D (0, −3, −7) between the reference value “50” and the reference value “50” are obtained.
[0045]
Next, the weighting coefficient y is multiplied by the determination value B from the second determination circuit 6 in the selected areas A2, A3, and A4. For example, as shown in FIG. 7, the weighting coefficient is the following weighting function y when the determination value B is the horizontal axis, the weighting coefficient y is the vertical axis, and the threshold value of the determination value B is the threshold value c.
y = −B / c + 1 (when judgment value B is equal to or less than threshold value c)
y = 0 (when judgment value B is larger than threshold c)
In the present embodiment, when the threshold value c = “10”, the weight coefficients y of the determination values B (0, 9, 49) of the areas A2, A3, A4 are (1, 0.1, 0), respectively.
[0046]
Next, the difference value D (0, -3, -7) and the weighting factor y (1, 0.1, 0) are multiplied for each region A2, A3, A4, and the weighted difference value WD (0x1, 0). −3 × 0.1, −7 × 0).
[0047]
Next, the weighted difference value WD is added to the reference value 50 except for the weighted difference value of the area A2 that is the reference value, and rounded off to (50 + (− 0.3) + 0 = 49.7) to obtain “50 Is output as an evaluation value.
[0048]
In the second embodiment and the first embodiment, the noise reduction effect by taking the average value does not change, but as the average value increases, the sharpness may be lost. The circuit has an effect of reducing noise to some extent and can realize an operation with little reduction in sharpness.
[0049]
In the above embodiment, the preferred example of the present invention has been described, but the present invention is not limited to this.
For example, in the embodiment described above, the line memory 2 is used to store and delay the video signal or image signal supplied from the signal input unit 1 for the horizontal scanning period. However, as shown in FIG. 2A may be used. In this case, by using the field memory 2A, the small area of FIG. 3 can be arbitrarily taken within one screen of the video signal or the image signal, so that the average value output circuit 3, the first determination circuit 4, the reference The value output circuit 5, the second determination circuit 6, and the evaluation value output circuit 7 can all be performed by software such as a microcomputer, so that a compact device can be obtained. Therefore, it can be used in portable display devices such as digital cameras (imaging devices), portable information devices, and portable communication devices.
[0050]
【The invention's effect】
As described above, according to the first aspect of the present invention, the first determination means determines whether each small region is flat or has an edge, and even if there are a plurality of flat portions, the second determination is made. Since the presence of a plurality of edges can be determined by the determination means and the evaluation value output means, the noise reduction effect can be enhanced even in a portion where there are no edges or a portion where a plurality of edges overlap.
[0051]
According to the second and third aspects of the present invention, since the presence of a plurality of edges can be determined by the evaluation value output means, there is an effect of reducing noise even in a portion with few edges or a portion where a plurality of edges overlap. .
[0052]
According to the fourth aspect of the present invention, there is an effect that the noise reduction effect is increased and the sharpness is less deteriorated even in a portion with few edges or a portion where a plurality of edges overlap.
[0053]
According to the fifth aspect of the present invention, the dispersion of each small region can be reliably obtained.
[Brief description of the drawings]
FIG. 1 is a block diagram of a noise reduction circuit according to a first embodiment of the present invention.
FIG. 2 is an explanatory diagram of schematic positions of small regions A1 to A9 including a target pixel in an n × n pixel range including the target pixel according to the first embodiment of the present invention.
FIG. 3 is an explanatory diagram of small regions A1 to A9 including a target pixel in an n × n pixel range including the target pixel according to the first embodiment of the present invention.
FIG. 4 is an explanatory diagram of the density of each pixel within a 5 × 5 pixel range of a target pixel and its neighboring pixels output from the line memory 2 according to the first embodiment of the present invention.
FIG. 5 shows output signals from the average value output circuit 3, the first determination circuit 4, the reference value output circuit 5, the second determination circuit 6, and the evaluation value output circuit 7 according to the first embodiment of the present invention. It is explanatory drawing which shows.
FIG. 6 shows output signals from the average value output circuit 3, the first determination circuit 4, the reference value output circuit 5, the second determination circuit 6, and the evaluation value output circuit 7 according to the second embodiment of the present invention. It is explanatory drawing which shows.
FIG. 7 is an explanatory diagram of a weighting function in the evaluation value output circuit according to the second embodiment of the present invention.
FIG. 8 is a block diagram of a noise reduction circuit according to another embodiment of the present invention.
[Explanation of symbols]
2 line memory 2A field memory 3 average value output circuit 4 first determination circuit 5 reference value output circuit 6 second determination circuit 7 evaluation value output circuit 10 field memories A1 to A9 small area e average values A and B determination value T Pixel of interest

Claims

An average value output that divides a processing area of a predetermined size into a plurality of small areas including a target pixel and neighboring pixels of the target pixel, and outputs a small area pixel average value that is an average value for a predetermined pixel value for each small area Means,
For each small area of the average value output means, a residual process for obtaining a residual between the small area pixel average value and a predetermined pixel value of each pixel, and a residual for obtaining an average value of the processing results after the residual process. A first determination means for outputting a first determination value indicating the dispersion of the pixel value for each small area from the difference average value processing;
Reference value output means for outputting, as a reference value, the small area pixel average value corresponding to the small area indicating the smallest processing result among the first determination values;
The variance of each small area with respect to the reference value is obtained by a residual process for obtaining a residual between the reference value and the average value of the small area pixels, and a residual average value process for obtaining an average value of the processing results after the residual process. Second determination means for outputting a second determination value to be indicated;
A noise reduction circuit having evaluation value output means for outputting an evaluation value related to a predetermined pixel value of the target pixel using the small region pixel average value, the first determination value, and the second determination value .

The evaluation value output means calculates the average value of the small region pixel average values of the corresponding small regions in which the first determination value is smaller than the first threshold and the second determination value is smaller than the second threshold, The noise reduction circuit according to claim 1, wherein the noise reduction circuit outputs the evaluation value.

The evaluation value output means obtains a small region pixel average value of a corresponding small region in which the first determination value is smaller than a first threshold and the second determination value is smaller than a second threshold;
The average value of the small area pixel average value obtained by removing the small area pixel average value of the small area that is the reference value from the small area pixel average value of the corresponding small area and the reference value is output as the evaluation value. The noise reduction circuit according to claim 1.

The evaluation value output means obtains a difference value between a small area pixel average value of a corresponding small area whose first determination value is smaller than a first threshold value and the reference value;
A process for obtaining a weighting coefficient of the second determination value corresponding to the corresponding small region with respect to a second threshold value;
The difference value and the weighting coefficient are subjected to a weighting process for obtaining a product for each small area,
The noise reduction circuit according to claim 1, wherein the evaluation value is obtained by adding the result of the weighting process to the reference value.

2. The first determination unit or the second determination unit performs variance processing, standard deviation processing, or absolute difference average processing by the residual processing and the residual average processing. The noise reduction circuit of thru | or 3.