JP4200745B2 - Signal processing apparatus and signal processing method - Google Patents

Description

【００３２】
一方、座標の回転は以下の数２式で表される。
【００３３】
【数２】

【００３４】
よって、図５に示す通り、ある画素について考えると移動５１および回転５２、５３は以下の数３式のように示される。
【００３５】
【数３】

【００３６】
次にＣｂ方向にＡ，Ｃｒ方向にＢ、拡大または圧縮すると、（ｘ”, ｙ”）は以下の数４式のように示される。
【００３７】
【数４】

【００３８】
以上より、ｘ" およびｙ" は以下の数５式のように示される。
【００３９】
【数５】

【００４０】
数５式を用いると、楕円状の距離e を移動、回転、拡大および圧縮した距離e ”は以下の数６式のように示すことができる。
【００４１】
【数６】

【００４２】
数６式では距離の計算に平方根の展開が必要となる。ここでは距離の比較に、距離を２乗した値を用いることで平方根の展開をなくすことができる。Ｃｂ方向およびＣｒ方向の拡大率（圧縮率）をＡ, Ｂとすると、数５式および数６式より楕円状距離計算ブロックは図６のように、ｘ、α、ｙ、β、Ａ，Ｂ、ｃｏｓθ、ｓｉｎθを入力として、減算器６１−1 〜６１−２、６３−１、加算器６３−２、６６、乗算器６２−１〜６２−４、６４−1 〜６４−２、６５−１〜６５−２で構成して、近時距離e ”² を出力することができることになる。
【００４３】
ここで、Ａを2 のａ乗、Ｂを２のｂ乗と限定すれば、拡大、圧縮のための乗算器６４−1 〜６４−２を削減することができ、楕円状の距離計算ブロックは図７に示すように、ａ、ｂのビットシフト演算を行うシフタ７４−１〜７４−２および簡単な乗算器７２−１〜７２−４、７５−１〜７５−２、減算器７１−1 〜７１−２、７３−１、加算器７３−２、７６、で実現できる。
【００４４】
さらに、回転角を３０度、６０度、１２０度、１５０度等に限定すれば、より簡単な乗算器およびシフタで回路を実現できる。例えば回転角が３０度の場合、回転行列は数７式となる。
【００４５】
【数７】

【００４６】
よって数５式は以下の数８式のようになる。
【００４７】
【数８】

【００４８】
図６と同様にＡを２のａ乗、Ｂを２のｂ乗とした場合、数８式より回転角が３０度の場合、楕円状の距離計算ブロックは図８のようにｘ、α、ｙ、β、√３を入力として、減算器８１−1 〜８１−２、８３−１、加算器８３−２、８６、乗算器８２−１〜８２−２、８５−１〜８５−２、ａ−１、ｂ−１のビットシフト演算を行うシフタ８４−1 〜８４−２で構成して、近時距離e ”² を出力するようにして実現できる。
【００４９】
回転角を３０ｎ度（ただしｎは３ｍではなく、ｍ, ｎは任意の整数）とすると、図７の入力のｃｏｓθ，ｓｉｎθは、±√３/ ２または±１/ ２となるため、楕円状距離計算ブロックはセレクタＳ１（９２−１）、セレクタＳ２（９２−２）を用いて、図９のように、ｘ、α、ｙ、β、√３、１、―√３、―１を入力として、セレクタＳ１（９２−１）、セレクタＳ２（９２−２）、減算器９１−1 〜９１−２、９４−１、加算器９４−２、９７、乗算器９３−１〜９３−４、９６−１〜９６−２、ａ−１、ｂ−１のビットシフト演算を行うシフタ９５−1 〜９５−２で構成して、近時距離e ”² を出力するようにして実現することが可能である。セレクタＳ１（９２−１）は図９のｃｏｓθの値を出力するように動作し、セレクタＳ２（９２−２）は図９のｓｉｎθの値を出力するように動作する。
【００５０】
ここで、図９に含まれるセレクタＳ１（９２−１）およびセレクタＳ２（９２−２）の動作について説明する。
図９は回転角を３０ｎ度（ただしｎは３ｍと等しくなく、ｍ, ｎは任意の整数である）とした場合の楕円状の距離計算ブロックであり、セレクタＳ１（９２−１）は図７のｃｏｓθ×２の値を出力するように動作し、セレクタＳ２（９２−２）は図７のｓｉｎθ×２の値を出力するように動作します。ここでは、後段の、ａ−１、ｂ−１のビットシフト演算を行うシフタ９５−1 〜９５−２でビットシフトしているためである。
【００５１】
具体的には、以下の表１に示すように、セレクタＳ１（９２−１）およびセレクタＳ２（９２−２）は、−√３, −１, １, √３の４つの値の中からｎの値によって１つの値を選択して出力する。
【００５２】
【表１】

【００５３】
さらに、図９において、図１０に示すセレクタＳ’（１０２−１，１０２−２）を用いると図１０のように乗算器９３−１〜９３−４を削減して、ｘ、α、ｙ、β、√３を入力として、セレクタＳ’（１０２−１，１０２−２）、減算器１０１−1 〜１０１−２、１０４−１、加算器１０４−２、１０７、乗算器１０３−１〜１０３−２、１０６−１〜１０６−２、ａ−１、ｂ−１のビットシフト演算を行うシフタ１０５−1 〜１０５−２で構成して、近時距離e ”² を出力するようにして構成することが可能である。
【００５４】
ただし、セレクタＳ‘（１０２−１，１０２−２）は図１０に示すｓｉｇ１〜ｓｉｇ４を以下の値、具体的には、sig1: ｘ―α、sig2: ｙ―β、sig3: √３（ｘ―α）、sig4: √３（ｙ―β）に設定した場合に表２に示すｎの値によって表２のような動作をするセレクタである。
【００５５】
【表２】

【００５６】
また、ｎ= ３ｍの場合には、楕円状距離計算ブロックは図１１のように、セレクタＳ’（１０２−１，１０２−２）、乗算器１０３−１〜１０３−２、減算器１０４−１、加算器１０４−２、を削減して、ｘ、α、ｙ、βを入力として、減算器１１１−1 〜１１１−２、ａ、ｂのビットシフト演算を行うシフタ１１２−1 〜１１２−２、１１３−１〜１１３−２乗算器、加算器１１４で構成して、近時距離e ”² を出力するようにして実現することが可能である。
【００５７】
以上に示した図６〜図１１の楕円状距離計算ブロックを用いると、特定色領域検出回路は図１２のように実現できる。図１２において、図６〜図１１の楕円状距離計算ブロック１２１は、画素情報Ｐおよび楕円領域情報Ｉから近時距離e ”² を算出し、乗算器１２２は抽出半径ｒの2 乗ｒ² を算出し、比較器（ｃｏｍｐ）１２３は近時距離e ”² と抽出半径Ｒの２乗ｒ² とを比較して、判定フラグＦを出力する。
【００５８】
ただし、図１２の楕円領域情報Ｉは色差信号Ｃｂ, Ｃｒ方向の拡大率およびＣ１の中心座標であり、比較器（ｃｏｍｐ）１２３のブロックは近時距離e ”² と抽出半径ｒの２乗ｒ² との値の大小を比較し、抽出半径ｒ² より楕円状距離e ”² の方が小さければ１を出力し、抽出半径ｒ² より楕円状距離e ”² の方が大きければ０を出力するブロックである。
【００５９】
本実施の形態では、楕円状の距離計算をし、距離比較をする際に平方根の展開をさけるために距離を２乗した値を比較に用いたが、必ずしも距離を用いて比較する必要はなく、距離に比例する値を使えばかまわない。よって、距離計算には例えば以下の数９式で示される１６角形近似（特開２００２−２１６１３６号公報参照）などの手法を用いてもかまわない。
【００６０】
【数９】

【００６１】
つぎに、第２のノイズ削減技術について説明する。
デジタル画像中に含まれるノイズを削減させる際に、例えば肌色のような特定色の画像に対してはフィルタのろ過作用を強めにかけてディテールを失わせることによって皺や肌荒れ等の肌領域のトラブルを目立たせにくくすることが考えられる。この場合特定色領域と特定色領域以外の境界部分で処理を急激に変化させると違和感のある画像となってしまうという問題がある。
【００６２】
そこで、注目画素のフィルタのろ過作用を変化させるためにフィルタ係数の重み付けを周辺画素の色情報を用いて行うことによって、特定色領域には強いろ過作用に対応する係数のフィルタをかけ、特定色以外の領域には弱いろ過作用に対応する係数のフィルタをかけ、さらに、特定色領域と特定色以外の領域との境界でノイズ削減レベルをなだらかに変化させることが可能となる。
【００６３】
ノイズの削減には、図１３Ａまたは図１３Ｂのような２次元のタップ係数をもつディジタルフィルタを用いる。図１３Ａまたは図１３Ｂのαが注目画素のフィルタ係数であり、注目画素の周辺画素が全て特定色であった場合にはαは最小値になり、最もフィルタのろ過作用が強くかかる。図１３には３×３フィルタと５×５フィルタを示したが、同等の特性が得られるものであれば、フィルタの係数、大きさは問わない。
【００６４】
図１４にフィルタ係数の導出のための特性図を示す。各特性Ｌ１、Ｌ２、Ｌ３について、縦軸に注目画素のフィルタ係数αを示し、横軸に周辺画素の特定色の数（ｃｏｕｎｔ）を示す。
【００６５】
図１４において、周辺画素が全て特定色でない場合には縦軸に示すフィルタ係数αはＦ＿ｍａｘになり、周辺画素が全て特定色の場合にはフィルタ係数はＦ＿ｍｉｎとなる。
【００６６】
図１４より横軸に示すように周辺画素の特定色の個数をカウント（ｃｏｕｎｔ）とすると、縦軸に示す注目画素のフィルタ係数αは横軸に示すカウント（ｃｏｕｎｔ）の値に依存するため、以下の数１０式のように表すことができる。
【００６７】
【数１０】

【００６８】
ここで、横軸に示す周辺画素の特定色の個数のカウント（ｃｏｕｎｔ）の値から注目画素のフィルタ係数αを求めるフィルタ係数算出ブロックをf ＿ｎｒ（ｃｏｕｎｔ）ブロック１５２とすると、ノイズ削減ブロックは図１５のように、上述した図３に示す特定色領域検出回路から出力される肌色等の特定色領域の検出を判定する特定色判定フラグＦ、輝度信号Ｙ、色差信号Ｃｒ、Ｃｂを、垂直方向に同時化するディレーライン（ＤＬ）１５１、特定色判定フラグＦに基づいて周辺画素の特定色の個数のカウント（ｃｏｕｎｔ）の値から注目画素のフィルタ係数αを求めるフィルタ係数算出ブロックのf ＿ｎｒ（ｃｏｕｎｔ）ブロック１５２、注目画素のフィルタ係数αの２次元のタップ係数により輝度信号ｙに対してろ過作用を施す２次元ディジタルフィルタ１５３で構成して、実現することが可能である。
【００６９】
例えば、図１４中の特性L ２を例にとって考えると上述した数１０式は以下の数１１式のように示すことができる。
【００７０】
【数１１】

【００７１】
数１１式で２項目の分母のＦｉｌｔｅｒ＿ｓｉｚｅはフィルタの要素数であり、ｃｏｕｎｔは周辺画素の特定色の個数である。数１１式をハードウエアで実現するとノイズ削減ブロックは図１６のように、上述した図３に示す特定色領域検出回路から出力される肌色等の特定色領域の検出を判定する特定色判定フラグＦ、輝度信号Ｙ、色差信号Ｃｒ、Ｃｂを、垂直方向に同時化するディレーライン（ＤＬ）１６１、特定色判定フラグＦに基づいて周辺画素の特定色の個数を加算して周辺画素の特定色個数ｘを出力する加算器１６４、Ｆ＿ｍａｘからＦ＿ｍｉｎを減算して減算値を出力する減算器１６２、減算値をＦｉｌｔｅｒ＿ｓｉｚｅで除算して除算値を出力する除算器１６３、周辺画素の特定色個数ｘと除算値とを乗算して乗算値を出力する乗算器１６５、乗算値とＦ＿ｍａｘを加算して注目画素のフィルタ係数αを出力する加算器１６６、注目画素のフィルタ係数αの２次元のタップ係数により輝度信号ｙに対してろ過作用を施す２次元ディジタルフィルタ１６７で構成して、実現することが可能でなる。しかし、図１６のノイズ削減ブロックでは除算器１６３を使っているためハードウェア規模が大きくなってしまう。
【００７２】
そこで、例えば、３×３フィルタを用いてＦ＿ｍａｘの値を１０、Ｆ＿ｍｉｎの値を１とした場合について考える。
このとき数１１式は以下の数１２式のようになる。
【００７３】
【数１２】

【００７４】
フィルタ係数αが数１２式で表される場合、ノイズ削減ブロックは図１７のように、上述した図３に示す特定色領域検出回路から出力される肌色等の特定色領域の検出を判定する特定色判定フラグＦ、輝度信号Ｙ、色差信号Ｃｒ、Ｃｂを、垂直方向に同時化するディレーライン（ＤＬ）１７１、特定色判定フラグＦに基づいて周辺画素の特定色の個数を加算して周辺画素の特定色個数ｘを出力する加算器１７２、周辺画素の特定色個数ｘからＦ＿ｍａｘを減算して注目画素のフィルタ係数αを出力する減算器１７３、注目画素のフィルタ係数αの２次元のタップ係数により輝度信号ｙに対してろ過作用を施す２次元ディジタルフィルタ１７４で構成して、実現することができ、これにより除算器１６３をなくし、簡単な回路構成とすることが可能である。
【００７５】
ここでは例として回路構成を簡単にするため、数１２式について説明したが、フィルタ係数αを求める数式は図１４の特性Ｌ１〜Ｌ３のように数１０式で示される一般的な式であればかまわない。
例えば、以下の数１３式のような式でもかまわない。
【００７６】
【数１３】

【００７７】
ただし、αｍａｘは周辺画素が全て特定色でなかった場合のαの値とする。
【００７８】
さらに、第３の特定色領域と特定色以外の領域の境界部分でゲインがなめらかに変化するアパーチャ補正技術について説明する。
デジタル画像信号処理では、画像の解像感、解像度を補正し、見映えを良くするためにアパーチャ補正（f 特補正( 周波数特性補正) ）を行うことが多い。
【００７９】
例えば、図１８のように、ディレーライン（ＤＬ）１８１により入力信号の輝度信号Ｙ、色差信号Ｃｒ、Ｃｂを、垂直方向に同時化する画像本線１８４に対して、ハイパスフィルタ( ＨＰＦ) １８２により輝度信号Ｙ、色差信号Ｃｒ、Ｃｂから画像の高域周波数成分１８５を抜き出して、加算器１８３により再び画像本線１８４の輝度信号Ｙ、色差信号Ｃｒ、Ｃｂに足す手法が知られている。
【００８０】
しかし、肌色のような特定色領域では、ディテールを失わせることによって皺や肌荒れ等の肌領域のトラブルを目立たせにくくするために、アパーチャ補正信号を小さくする必要がある。
【００８１】
しかし、全領域に対してアパーチャ補正信号を小さくしてしまうと、ボケた画像になってしまうので特定色領域のアパーチャ補正信号のみを小さくして、特定色以外の領域のアパーチャ補正信号は変更しないことが望ましい。さらに、出力画像に違和感を与えないためには、特定色領域と特定色以外の領域でアパーチャ補正信号がなめらかに変化している必要がある。
【００８２】
そこで、図１９のように、ディレーライン（ＤＬ）１９１により入力信号の輝度信号Ｙ、色差信号Ｃｒ、Ｃｂを、垂直方向に同時化し、垂直ローパスフィルタ( Ｖ−ＬＰＦ) １９２により輝度信号Ｙの垂直方向の低域周波数成分を抜き出し、水平バンドパスフィルタ( Ｈ−ＢＰＦ) １９４により輝度信号Ｙの水平方向の所定帯域周波数成分を抜き出し、垂直バンドパスフィルタ( Ｖ−ＢＰＦ) １９３により輝度信号Ｙの垂直方向の所定帯域周波数成分を抜き出し、水平ローパスフィルタ( Ｈ−ＬＰＦ) １９５により輝度信号Ｙの水平方向の低域周波数成分を抜き出し、加算器１９７によりアパーチャ補正信号を出力し、特定色判定フラグＦの注目画素周辺の色情報をもとに、ゲイン算出ブロック１９６により周辺画素の特定色の個数のカウント（ｃｏｕｎｔ）の値から注目画素のゲインを求め、アパーチャ補正信号にゲインを与え、乗算器１９８によりアパーチャ補正信号にゲインを乗算する。
【００８３】
周辺画素の特定色の個数をカウント（ｃｏｕｎｔ）とした場合に、注目画素のアパーチャ補正信号のゲインは、カウント（ｃｏｕｎｔ）の値に依存するため、数１４式のように表すことができる。
【００８４】
【数１４】

【００８５】
数１４式をハードウエアで実現したものが、図１９に示したゲイン算出ブロック１９６となる。
【００８６】
ここで、図２０のアパーチャ補正信号のゲインの導出図で示す特性Ｌ１、Ｌ２、Ｌ３のように、横軸の周辺画素の特定色の個数をカウント（ｃｏｕｎｔ）としたとき、縦軸のゲインをなめらかに変化させることで、特定色領域のみのアパーチャ補正信号を小さくし、特定色以外の領域との境界でアパーチャ補正信号をなめらかに変化させることが可能となる。
【００８７】
ここで、図２０の特性Ｌ２を例にとって考えると、アパーチャ補正信号のゲインを算出する式は以下の数１５式のようになる。
【００８８】
【数１５】

【００８９】
数１５式をハードウエアで実現すると、図１９に示したゲイン算出ブロック１９６は図２１のように、特定色判定フラグＦに基づいて周辺画素の特定色の個数を加算して周辺画素の特定色個数ｘを出力する加算器２１１、Ｇ＿ｍａｘからＧ＿ｍｉｎを減算して減算値を出力する減算器２１２、減算値をＦｉｌｔｅｒ＿ｓｉｚｅで除算して除算値を出力する除算器２１３、周辺画素の特定色個数ｘと除算値とを乗算して乗算値を出力する乗算器２１４、乗算値とＧ＿ｍａｘを加算して注目画素のアパーチャ補正信号のゲインＧを出力する加算器２１５で構成して、実現することが可能である。この図２１では除算器２１３が必要となるが、ここで、図２０のＸ＿ｖａｌｕｅを２のｘ’乗とすると、数１５式は数１６式のように変形することができる。
【００９０】
【数１６】

【００９１】
Ｘ＿ｖａｌｕｅは２のｘ' 乗であるので、数１６式中の除算はｘ' のビットシフトを行うシフタで実現できることになる。数１６式中の除算をｘ' のビットシフトを行うシフタで実現した場合のゲイン算出ブロックを図２２に示すと、特定色判定フラグＦに基づいて周辺画素の特定色の個数を加算して周辺画素の特定色個数ｘを出力する加算器２２１、Ｇ＿ｍａｘのｘ' のビットシフトを行ってビットシフト値を出力するシフタ２２４、周辺画素の特定色個数ｘとＧ＿ｍａｘのｘ' のビットシフト値を乗算して乗算値を出力する乗算器２２２、乗算値とＧ＿ｍａｘを加算して注目画素のアパーチャ補正信号のゲインＧを出力する加算器２２３で構成して、実現することが可能である。
【００９２】
アパーチャ補正信号のゲイン導出についてここでは例として数１６式を説明したが、図１５の特性Ｌ１〜Ｌ３のように数１４式で示される一般的な式であればかまわない。
例えば、以下の数１７式のような式でもかまわない。
【００９３】
【数１７】

【００９４】
ただし、数１７式中のｇｍａｘは周辺画素が全て特定色でなかった場合のｇ（ゲイン）の値とする。
【００９５】
以上より、アパーチャ補正ブロックは図２３のように、ディレーライン（ＤＬ）２３１により入力信号の輝度信号Ｙ、色差信号Ｃｒ、Ｃｂを、垂直方向に同時化し、垂直ローパスフィルタ( Ｖ−ＬＰＦ) ２３２により輝度信号Ｙの垂直方向の低域周波数成分を抜き出し、水平バンドパスフィルタ( Ｈ−ＢＰＦ) ２３４により輝度信号Ｙの水平方向の所定帯域周波数成分を抜き出し、垂直バンドパスフィルタ( Ｖ−ＢＰＦ) ２３３により輝度信号Ｙの垂直方向の所定帯域周波数成分を抜き出し、水平ローパスフィルタ( Ｈ−ＬＰＦ) ２３５により輝度信号Ｙの水平方向の低域周波数成分を抜き出し、加算器２３６によりアパーチャ補正信号を出力し、特定色判定フラグＦの注目画素周辺の色情報をもとに、ゲイン算出ブロック２３７により周辺画素の特定色の個数のカウント（ｃｏｕｎｔ）の値から注目画素のゲインを求め、アパーチャ補正信号にゲインを与え、乗算器２３８によりアパーチャ補正信号にゲインを乗算し、加算器２３９によりこの乗算値を輝度信号ｙ、色差信号Ｃｒ、Ｃｂに加算する。
【００９６】
図２３中の「Ｖ−ＬＰＦ」２３２は垂直方向のローパスフィルタであり、「Ｖ−ＢＰＦ」２３３は垂直方向のバンドパスフィルタである。同様に「 Ｈ−ＬＰＦ」 ２３５、「 Ｈ−ＢＰＦ」 ２３４はそれぞれ水平方向のローパスフィルタ、水平方向のバンドパスフィルタを示す。なお、図示はしないが、乗算器２３８の後段にコアリング（ｃｏｒｉｎｇ）部として微少振幅をノイズとみなし、ノイズとみなした部分のアパーチャ信号は０にするという処理をするブロックを付加してもよい。
【００９７】
また、第４に図１２に示した特定色領域検出回路における特定色領域検出において比較器（ｃｏｍｐ）１２３部分の判定フラグＦの出力を距離に応じた多ビットの信号にした場合について説明する。
【００９８】
特定色領域検出において、特定色領域に含まれているかどうかの判定結果を多ビットで表現することにより、図２４に示すように「特定色らしさ」を示すことが可能となる。ある画素が特定色領域の中心に近ければ近いほど、特定色に近いということになる。
【００９９】
例えば、図２４で判定フラグＦの特定色らしさを２ビットで表現すると仮定した場合について説明する。中心に最も近い領域に含まれる場合には重み付けを３とし、中心から次に近い領域に含まれる場合には２とし、図２２中の楕円状領域の中で最も外側の領域に含まれる場合を1 とし、楕円状領域に含まれていない場合には０とする。このようにある画素が特定色領域の中心に近ければ近いほど大きい値を割り当てることで、判定結果の値の大きさを用いてどのくらい楕円状領域の中心に近いのかを判定することが可能となる。
【０１００】
例として、特定色らしさを2 ビットで表現する場合について説明したが、特定色らしさを表現するには何ビットでもかまわない。
【０１０１】
特定色領域検出ブロックの判定結果を多ビットで表現すると、1 ビットで表現した場合に比べてノイズ削減部で注目画素のフィルタ係数がなめらかに変化し、アパーチャ補正部でアパーチャ補正信号のゲインがなめらかに切り替わることになる。そのため、図１２や図１５等に示した特定色判定フラグＦが多ビットになるが、その他のブロックの構成は変更しなくても、特定色判定結果を1 ビットで表現した場合に比べて、特定色領域と特定色以外の領域の境界でよりなめらかな処理の実現が可能となる。
【０１０２】
上述した本実施の形態によれば、例えば肌色などの特定色のノイズを削減することができ、ノイズがほとんどない領域に対しても、故意にノイズ削減（ＮＲ）を強くかけアパーチャ補正信号を弱くしてディテールを失わせることによって、例えば、肌色領域ならば「にきび」や「しみ」といった肌のトラブルを目立ちにくくし、綺麗な肌を再現することが可能となる。しかも、他の色領域については変化させないようにすることができる。
【０１０３】
上述した本実施の形態に限らず、本発明の特許請求の範囲を逸脱しない範囲で適宜他の構成をとりうることができることは言うまでもない。
【０１０４】
【発明の効果】
この発明の信号処理装置は、撮像手段により撮像された画像データの輝度信号および色差信号に基づき、色差平面で各画素の色差値が特定色領域に含まれるか否かの判定結果を、特定色領域の中心に近いほど値の大きい多ビットの信号で表現する特定色領域検出手段と、特定色領域検出手段の判定結果に基づき、注目画素のノイズを削減するフィルタのフィルタ係数を、周辺画素についての多ビットの信号のカウント値が大きいほど小さくするノイズ削減手段とを備え、特定色領域における信号処理を行うようにしたので、画像処理を行う際に、肌色や空色等の特定色領域に含まれるか否かの判定結果を多ビットで表現することにより、１ビットで表現した場合に比べて注目画素のフィルタ係数がなめらかに変化する。これにより、特定色領域と特定色領域以外の境界部分で処理をなめらかに変化させ、画質に違和感を与えないようにノイズ削減を行うことができるので、例えば肌色などの特定色のノイズを削減して滑らかな画像を生成して、画像品質を向上させることができるという効果を奏する。
【０１０５】
また、この発明の信号処理装置は、上述において、特定色領域検出手段は、色差平面で楕円状の特定色領域に対する各画素の距離を移動、回転させた後に縦横を圧縮・拡大した距離の２乗の値と、色差平面で楕円状の特定色領域を移動、回転させた後に縦横を圧縮・拡大した円の半径の２乗の値とを比較して判定を行うので、特定色領域検出手段を簡単な回路で実現できるという効果を奏する。
【０１０６】
また、この発明の信号処理装置は、上述において、特定色領域検出手段は、圧縮・拡大の比率を２の累乗に限定するか、回転角度を３０度、６０度、１２０度、１５０度に限定するかの少なくとも一方の限定をして計算を行うので、特定色領域検出手段をより簡単な回路で実現できるという効果を奏する。
【０１０７】
また、この発明の信号処理装置は、撮像手段により撮像された画像データの輝度信号および色差信号に基づき、色差平面で各画素の色差値が特定色領域に含まれるか否かの判定結果を、特定色領域の中心に近いほど値の大きい多ビットの信号で表現する特定色領域検出手段と、特定色領域検出手段の判定結果に基づき、注目画素に対するアパーチャ補正信号のゲインを、周辺画素についてのこの多ビットの信号のカウント値が大きいほど小さくするアパーチャ補正手段とを備え、特定色領域における信号処理を行うようにしたので、画像処理を行う際に、肌色や空色等の特定色領域に含まれるか否かの判定結果を多ビットで表現することにより、１ビットで表現した場合に比べてアパーチャ補正信号のゲインがなめらかに切り替わる。これにより、特定色領域と特定色領域以外の境界部分で処理をなめらかに変化させ、画質に違和感を与えないようにアパーチャ補正を行うことができるので、例えば肌色などの特定色をアパーチャ補正して滑らかな画像を生成して、画像品質を向上させることができるという効果を奏する。
【０１０８】
また、この発明の信号処理装置は、上述において、特定色領域検出手段は、色差平面で楕円状の特定色領域に対する各画素の距離を移動、回転させた後に縦横を圧縮・拡大した距離の２乗の値と、色差平面で楕円状の特定色領域を移動、回転させた後に縦横を圧縮・拡大した円の半径の２乗の値とを比較して判定を行うので、特定色領域検出手段を簡単な回路で実現できるという効果を奏する。
【０１０９】
また、この発明の信号処理装置は、上述において、特定色領域検出手段は、圧縮・拡大の比率を２の累乗に限定するか、回転角度を３０度、６０度、１２０度、１５０度に限定するかの少なくとも一方の限定をして計算を行うので、特定色領域検出手段をより簡単な回路で実現できるという効果を奏する。
【０１１０】
また、この発明の信号処理方法は、撮像手段により撮像された画像データの輝度信号および色差信号に基づき、色差平面で各画素の色差値が特定色領域に含まれるか否かの判定結果を、特定色領域の中心に近いほど値の大きい多ビットの信号で表現する特定色領域検出ステップと、特定色領域検出ステップの判定結果に基づき、注目画素のノイズを削減するフィルタのフィルタ係数を、周辺画素についてのこの多ビットの信号のカウント値が大きいほど小さくするノイズ削減ステップとを備え、特定色領域における信号処理を行うようにしたので、画像処理を行う際に、肌色や空色等の特定色領域に含まれるか否かの判定結果を多ビットで表現することにより、１ビットで表現した場合に比べて注目画素のフィルタ係数がなめらかに変化する。これにより、特定色領域と特定色領域以外の境界部分で処理をなめらかに変化させ、画質に違和感を与えないようにノイズ削減を行うことができるので、例えば肌色などの特定色のノイズを削減して滑らかな画像を生成して、画像品質を向上させることができるという効果を奏する。
【０１１１】
また、この発明の信号処理方法は、上述において、特定色領域検出ステップは、色差平面で楕円状の特定色領域に対する各画素の距離を移動、回転させた後に縦横を圧縮・拡大した距離の２乗の値と、色差平面で楕円状の特定色領域を移動、回転させた後に縦横を圧縮・拡大した円の半径の２乗の値とを比較して判定を行うので、特定色領域検出ステップを行うブロックを簡単な回路で実現できるという効果を奏する。
【０１１２】
また、この発明の信号処理方法は、上述において、特定色領域ステップは、圧縮・拡大の比率を２の累乗に限定するか、回転角度を３０度、６０度、１２０度、１５０度に限定するかの少なくとも一方の限定をして計算を行うので、特定色領域検出ステップを行うブロックをより簡単な回路で実現できるという効果を奏する。
【０１１３】
また、この発明の信号処理方法は、撮像手段により撮像された画像データの輝度信号および色差信号に基づき、色差平面で各画素の色差値が特定色領域に含まれるか否かの判定結果を、特定色領域の中心に近いほど値の大きい多ビットの信号で表現する特定色領域検出ステップと、特定色領域検出ステップの判定結果に基づき、注目画素に対するアパーチャ補正信号のゲインを、周辺画素についての多ビットの信号のカウント値が大きいほど小さくするアパーチャ補正ステップとを備え、特定色領域における信号処理を行うようにしたので、画像処理を行う際に、肌色や空色等の特定色領域に含まれるか否かの判定結果を多ビットで表現することにより、１ビットで表現した場合に比べてアパーチャ補正信号のゲインがなめらかに切り替わる。これにより、特定色領域と特定色領域以外の境界部分で処理をなめらかに変化させ、画質に違和感を与えないようにアパーチャ補正を行うことができるので、例えば肌色などの特定色をアパーチャ補正して滑らかな画像を生成して、画像品質を向上させることができるという効果を奏する。
【０１１４】
また、この発明の信号処理方法は、上述において、特定色領域検出ステップは、色差平面で楕円状の特定色領域に対する各画素の距離を移動、回転させた後に縦横を圧縮・拡大した距離の２乗の値と、色差平面で楕円状の特定色領域を移動、回転させた後に縦横を圧縮・拡大した円の半径の２乗の値とを比較して判定を行うので、特定色領域検出ステップを行うブロックを簡単な回路で実現できるという効果を奏する。
【０１１５】
また、この発明の信号処理方法は、上述において、特定色領域ステップは、圧縮・拡大の比率を２の累乗に限定するか、回転角度を３０度、６０度、１２０度、１５０度に限定するかの少なくとも一方の限定をして計算を行うので、特定色領域検出ステップを行うブロックをより簡単な回路で実現できるという効果を奏する。
【図面の簡単な説明】
【図１】デジタルスチルカメラのブロック図である。
【図２】カメラシステムＬＳＩのブロック図である。
【図３】本発明の実施の形態に適用されるカメラ信号処理部のブロック図である。
【図４】特定色（肌色）の色差分布を示す図である。
【図５】楕円状の距離計算手順を示す図である。
【図６】楕円状距離計算ブロックを示す図である。
【図７】拡大、圧縮をシフトで実現した楕円状の距離計算ブロックを示す図である。
【図８】回転角が３０度の場合の楕円状の距離計算ブロックを示す図である。
【図９】回転角を3 ０ｎ度（ただしｎ≠３ｍ：ｍ, ｎは任意の整数）とした場合の楕円状の距離計算ブロックを示す図である。
【図１０】セレクタＳ’を用いた楕円状の距離計算ブロックを示す図である。
【図１１】ｎ= ３ｍの場合の楕円状距離計算ブロックを示す図である。
【図１２】特定色領域検出回路を示す図である。
【図１３】ＮＲフィルタを示す図であり、図１３Ａは３×３フィルタ、図１３Ｂは５×５フィルタである。
【図１４】フィルタ係数αの導出を示す図である。
【図１５】ノイズ削減ブロック（１）を示す図である。
【図１６】ノイズ削減ブロック（２）を示す図である。
【図１７】ノイズ削減ブロック（３）を示す図である。
【図１８】一般的なアパーチャ補正を示す図である。
【図１９】アパーチャ補正信号の導出を示す図である。
【図２０】アパーチャ補正信号のゲインの導出を示す図である。
【図２１】ゲイン算出ブロック（１）を示す図である。
【図２２】ゲイン算出ブロック（２）を示す図である。
【図２３】アパーチャ補正ブロックを示す図である。
【図２４】特定色らしさを示す図である。
【符号の説明】
１……レンズ、２……絞り、３……シャッター、４……画像センサー、５……ＴＧ（タイミングジェネレーター）、６……フロントエンド、７……カメラシステムＬＳＩ、８……画像メモリ、９……マイコン、１０……外部記憶媒体、１１……画像モニター、２１……画像検波部、２２……カメラ信号処理部、２３……メモリＩ／Ｆ、２４……メモリコントローラ、２５……マイコンＩ／Ｆ、２６……画像圧縮・解凍部、２７……モニターＩ／Ｆ、３１……特定色領域検出部、３２……ＤＬ（ディレイライン）、３３……ＮＲ（ノイズリダクション）部、３４……ＡＰ（アパーチャ―）補正部、４１……楕円状領域( 肌色) 、４２……楕円状領域( 空色) 、５１……移動、５２、５３……回転、５４……拡大・圧縮、１２１……楕円状距離計算ブロック[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a signal processing apparatus and a signal processing method for performing signal processing such as noise removal and aperture correction on image data captured by an imaging unit, for example.
[0002]
[Prior art]
In recent years, as the size and performance of digital cameras have become smaller, the number of pixels that can image a subject with a CCD (Charge Coupled Device) as an imaging device has increased, and the cell size that constitutes one pixel has therefore decreased. The ratio of noise to the CCD sensor output signal is large.
[0003]
As a result, techniques such as noise removal based on the dark current of the CCD and aperture correction are required for specific colors such as skin color and sky blue. In the field of digital color images, detection of specific colors such as skin color, and Techniques such as noise removal and aperture correction for the detected specific color region are generally well known (Patent Documents 1 to 6).
[0004]
[Patent Document 1]
JP-A-8-210916
[Patent Document 2]
Japanese Patent Laid-Open No. 11-146405
[Patent Document 3]
JP-A-8-210916
[Patent Document 4]
Japanese Patent Laid-Open No. 11-146405
[Patent Document 5]
JP 2000-152268 A
[Patent Document 6]
Japanese Patent Laid-Open No. 11-146405
[Patent Document 7]
Japanese Patent Laid-Open No. 10-293840
[0005]
[Problems to be solved by the invention]
As a conventional specific color detection technique described above, there is Patent Document 3, and a specific color is detected using the upper limit value and the lower limit value of each RGB signal level. However, this method requires a coordinate conversion table. There is an inconvenience that the storage capacity of the memory increases to detect a specific color.
[0006]
Further, Patent Document 4 also has a disadvantage that the hardware scale increases because the color difference signal is written as an address in a table in the memory.
[0007]
Further, as a noise removal technique, there is Patent Document 5, which is a technique of changing the noise determination level depending on the color and performing noise removal using the noise slice level. Therefore, the boundary between the specific color region and the region other than the specific color is used. Although the noise level changes abruptly at the portion, in the present invention, the noise level is sharply changed at the boundary portion between the specific color region and the region other than the specific color by performing noise removal based on pixel information around the specific color. The challenge is to keep it from changing.
[0008]
Further, there is Patent Document 6 as an aperture correction technique, and the frequency characteristic is changed without changing the gain of the aperture correction signal using BPF (Band Pass Filter) having different frequency characteristics, and therefore the gain of the aperture correction signal is set. Although it cannot be changed and the circuit configuration is complicated, in the present invention, since the gain of the aperture correction signal is calculated based on information around the target pixel for a specific color, an appropriate gain can be obtained with a simpler circuit. It is an object to generate an aperture correction signal. Patent Document 7 discloses a technique for normalizing a binarized face area of a captured image. In the present invention, a specific color is used for noise removal and aperture correction on a color difference plane based on a color difference signal. It is an object to detect a region.
[0009]
Therefore, the present invention has been made in view of such a point, and a specific color is detected without increasing the storage capacity of the memory, and the noise level is rapidly increased at the boundary portion between the specific color area and the area other than the specific color. It is an object of the present invention to provide a signal processing apparatus and a signal processing method capable of performing aperture correction with a simpler circuit while preventing the change.
[0010]
[Means for Solving the Problems]
The signal processing apparatus of the present invention is based on the luminance signal and color difference signal of image data. The Color difference value of each pixel in the color difference plane The result of determining whether or not the image is included in the specific color area is expressed by a multi-bit signal having a larger value as it is closer to the center of the specific color area Specific color area detection means and specific color area detection means Based on the determination result, the filter coefficient of the filter that reduces the noise of the target pixel is decreased as the count value of the multi-bit signal for the peripheral pixel is increased. And a noise reduction means for performing signal processing in a specific color region.
[0011]
Therefore, according to the present invention, the following operations are performed.
Specific color areas such as skin color and sky blue when performing image processing If the determination result of whether or not it is included is expressed in multiple bits, the filter coefficient of the pixel of interest changes more smoothly than when expressed in 1 bit. This Smoothly change the process at the boundary between the specific color area and the specific color area to reduce noise so that the image quality does not feel strange be able to .
[0012]
The signal processing apparatus of the present invention is based on the luminance signal and color difference signal of image data. The Color difference value of each pixel in the color difference plane The result of determining whether or not the image is included in the specific color area is expressed by a multi-bit signal having a larger value as it is closer to the center of the specific color area Specific color area detection means and specific color area detection means Based on the determination result, the gain of the aperture correction signal for the target pixel is decreased as the count value of the multi-bit signal for the peripheral pixel is increased. An aperture correction unit is provided to perform signal processing in a specific color region.
[0013]
Therefore, according to the present invention, the following operations are performed.
Specific color areas such as skin color and sky blue when performing image processing When the determination result of whether or not it is included is expressed by multiple bits, the gain of the aperture correction signal is smoothly switched as compared with the case where it is expressed by 1 bit. This Smoothly change the process at the boundary between the specific color area and the specific color area, and perform aperture correction so that the image quality does not feel strange be able to .
[0014]
The signal processing method of the present invention is based on the luminance signal and color difference signal of image data. The Color difference value of each pixel in the color difference plane The result of determining whether or not the image is included in the specific color area is expressed by a multi-bit signal having a larger value as it is closer to the center of the specific color area Specific color area detection step and specific color area detection step Based on the determination result, the filter coefficient of the filter that reduces the noise of the target pixel is decreased as the count value of the multi-bit signal for the peripheral pixel is increased. And a noise reduction step to perform signal processing in a specific color region.
[0015]
Therefore, according to the present invention, the following operations are performed.
Specific color areas such as skin color and sky blue when performing image processing If the determination result of whether or not it is included is expressed in multiple bits, the filter coefficient of the pixel of interest changes more smoothly than when expressed in 1 bit. This Smoothly change the process at the boundary between the specific color area and the specific color area to reduce noise so that the image quality does not feel strange be able to .
[0016]
The signal processing method of the present invention is based on the luminance signal and color difference signal of image data. The Color difference value of each pixel in the color difference plane The result of determining whether or not the image is included in the specific color area is expressed by a multi-bit signal having a larger value as it is closer to the center of the specific color area Specific color area detection step and specific color area detection step Based on the determination result, the gain of the aperture correction signal for the target pixel is decreased as the count value of the multi-bit signal for the peripheral pixel is increased. An aperture correction step, and performs signal processing in a specific color region.
[0017]
Therefore, according to the present invention, the following operations are performed.
Specific color areas such as skin color and sky blue when performing image processing When the determination result of whether or not it is included is expressed by multiple bits, the gain of the aperture correction signal is smoothly switched as compared with the case where it is expressed by 1 bit. This Smoothly change the process at the boundary between the specific color area and the specific color area, and perform aperture correction so that the image quality does not feel strange be able to .
[0018]
DETAILED DESCRIPTION OF THE INVENTION
Embodiments of the present invention will be described below with reference to the drawings as appropriate.
FIG. 1 shows a block diagram of a digital still camera.
First, the configuration of the digital still camera will be described.
In FIG. 1, a lens 1 collects reflected light of a subject and guides it to an image sensor 4. The diaphragm 2 controls the amount of light hitting the image sensor 4 by changing the area through which light passes. The shutter 3 controls exposure by blocking the passage of light. The diaphragm 2 described above may also function as the shutter 3. The image sensor 4 is an image sensor such as a CCD or a CMOS (Complementary Metal-Oxide Semiconductor) sensor, and converts light information of a subject into electrical information. A plurality of color filters are arranged on the sensor surface to identify colors.
[0019]
The front end (S / H (Sample and Hold) 6-1, GC (Gain Control) 6-2, A / D (Analog to Digital) 6-3) 6 portion converts the analog signal output from the image sensor 4 into a digital signal. The signal is converted into a signal, and internally correlated double sampling, gain control, and A / D conversion for removing noise components are performed. A timing generator (TG) 5 drives the image sensor 4 horizontally and vertically, and performs exposure control such as a high-speed / low-speed electronic shutter. The camera system LSI (Large Scale Integration) 7 performs various digital signal processing based on the digitized information from the image sensor 4 to generate a luminance signal and a color signal, compresses the data, and compresses the compressed data. Generate.
[0020]
The image memory 8 is a storage element for storing digital image data, and includes a DRAM (Dynamic Random Access Memory), an SDRAM (Synchronous Random Access Memory), and the like, and is used for digital signal processing in the camera system LSI 7. A microcomputer (hereinafter referred to as a microcomputer) 9 controls the entire camera block. Specifically, exposure control by the diaphragm 2, opening / closing control of the shutter 3, electronic shutter control at the TG 5, and front end 6 Controls such as gain control, various mode control of the camera system LSI 7 and parameter control are performed.
[0021]
The external storage medium 10 is a storage medium for storing compressed data, for example, JPEG (Japan Print Experts group) data, and is configured to be detachable from the digital still camera body and to be taken out and carried. The image monitor 11 is a monitor for confirming a captured image, and is an LCD (Charge Coupled Device) panel or the like.
[0022]
The operation of the digital still camera configured as described above will be described below.
FIG. 2 is a block diagram of the camera system LSI.
In FIG. 2, an image detection unit 21 performs detection processing of a camera-captured image serving as a reference for various camera controls such as auto focus (AF), auto exposure (AE), and auto white balance (AWB). A microcomputer I / F (interface) 25 performs a bus I / F with the microcomputer 9 that controls the camera system LSI 7, and exchanges control data and image data between the microcomputer 9 and the system LSI 7. The memory I / F 23 performs I / F with the image memory 8 used in signal processing in the system LSI 7 and transfers image data and compressed data.
[0023]
The memory controller 24 transfers image data between various signal processing blocks or between the signal processing block and the memory and controls a data path through which data flows. The image compression / decompression unit 26 compresses image data or decompresses compressed image data, and is, for example, a JPEG encoder. The monitor I / F 27 converts image data into various display formats for display on an external monitor or display. For example, the monitor I / F 27 is an NTSC encoder for displaying on an NTSC (National Television System Standard Committee) monitor. The camera signal processing unit 22 performs digital signal processing such as interpolation calculation, filter calculation, matrix calculation, luminance generation calculation, and color generation calculation based on the sensor image information digitized by the front end unit 6 to obtain luminance / color signals. An image signal consisting of
[0024]
In FIG. 1, the S / H 6-1, GC 6-2, A / D 6-3 part (front end 6) generates noise for the output signal of the image CCD sensor 4 that has received a drive pulse from the timing generator (TG) 5. Correlated double sampling to remove components, gain control, and A / D conversion are performed to convert an analog signal into a digital signal. Thereafter, the camera signal processing portion 22 of the camera system LSI 7 (FIG. 2) performs digital clamping, white balance, gamma correction, interpolation processing, and the like to generate RGB values and then convert them into luminance and color difference signals.
[0025]
The processing of the present application is included in the camera signal processing section 22, and the luminance and chrominance signals are input to perform specific color area detection, noise reduction, and aperture correction processing, and then the luminance and chrominance signals are output to the JPEG encoding circuit. A block diagram of the signal processing unit of the present application is shown in FIG. The specific color area detection circuit 31 outputs a specific color determination flag F for determining the detection of a specific color area such as skin color based on the luminance signal Y, color difference signals Cr and Cb, ellipse area information I, and extraction radius r information. To do. The delay line (DL) 32 synchronizes signals in the vertical direction. After that, noise is reduced based on peripheral pixel information by the noise reduction unit (NR) 33 based on the information of the luminance signal Y and the specific color determination flag F, and aperture correction is performed by the aperture correction unit (AP) 34. The corrected luminance signal Y is output to the subsequent JPEG encoding circuit.
[0026]
First, the specific color area detection technique will be described.
In general, the color difference distribution on the Cr and Cb planes by a color difference signal of a specific color such as skin color is as shown in FIG. 4. For example, an elliptical region (skin color) 41 or an elliptical region (sky blue) 42 is elliptical. Is distributed. For this reason, it has been confirmed through experiments that the detection of specific color regions is less likely to occur when an ellipse is used than when a circle is used. The specific color area can be detected by determining whether the color difference signal is included in the elliptical area (skin color) 41 or the elliptical area (sky blue) 42 in FIG.
[0027]
However, if the area detection by the ellipse is to be realized by hardware, the circuit scale increases, and thus a processing device is required. Therefore, we propose an elliptical distance calculation method that can be easily implemented by hardware.
[0028]
To determine whether or not the pixel (x, y) in the elliptical area C1 shown in FIG. 5 is included, the coordinate movement 51 of the certain pixel (x, y) and the elliptical area C1 is performed, and rotations 52 and 53 are enlarged. Alternatively, it can be realized by comparing values subjected to the compression 54 processing.
[0029]
Assuming that the center coordinates of the elliptical region C1 are (α, β), the radius of the circular region C2 (region extraction radius) is r, and C2 translates 51 and rotates 52 and 53, then C2 direction The circle is expanded or compressed by B in the A and Cr directions. That is, it is only necessary to determine whether or not the distance e ″ obtained by applying the processing of the movement 51, the rotations 52 and 53, the enlargement and the compression 54 to the elliptical distance e in FIG.
[0030]
The parallel movement 51 of a certain pixel can be expressed by the following equation (when α is moved in the x direction and β is moved in the y direction).
[0031]
[Expression 1]

[0032]
On the other hand, the rotation of coordinates is expressed by the following equation (2).
[0033]
[Expression 2]

[0034]
Therefore, as shown in FIG. 5, when a certain pixel is considered, the movement 51 and the rotations 52 and 53 are expressed by the following Equation 3.
[0035]
[Equation 3]

[0036]
Next, when A is expanded in the Cb direction and B is expanded or compressed in the Cr direction, (x ″, y ″) is expressed by the following equation (4).
[0037]
[Expression 4]

[0038]
From the above, x ″ and y ″ are expressed by the following equation (5).
[0039]
[Equation 5]

[0040]
Using Equation 5, the distance e ″ obtained by moving, rotating, expanding, and compressing the elliptical distance e 1 can be expressed by the following Equation 6.
[0041]
[Formula 6]

[0042]
In Equation 6, square root expansion is required for distance calculation. Here, the square root expansion can be eliminated by using a value obtained by squaring the distance in the comparison of distances. Assuming that the enlargement ratios (compression ratios) in the Cb direction and the Cr direction are A and B, the elliptical distance calculation block is expressed as x, α, y, β, A, B as shown in FIG. , Cos θ, and sin θ as inputs, subtracters 61-1 to 61-2 and 63-1, adders 63-2 and 66, multipliers 62-1 to 62-4, 64-1 to 64-2, and 65- 1-65-2, the short distance e " ² Can be output.
[0043]
Here, if A is limited to 2 to the power of a and B is limited to the power of 2 to b, the multipliers 64-1 to 64-2 for enlargement and compression can be reduced. As shown in FIG. 7, shifters 74-1 to 74-2 that perform bit shift operations of a and b, simple multipliers 72-1 to 72-4, 75-1 to 75-2, and a subtractor 71-1. ~ 71-2, 73-1, and adders 73-2, 76.
[0044]
Furthermore, if the rotation angle is limited to 30 degrees, 60 degrees, 120 degrees, 150 degrees, etc., a circuit can be realized with a simpler multiplier and shifter. For example, when the rotation angle is 30 degrees, the rotation matrix is expressed by Equation 7.
[0045]
[Expression 7]

[0046]
Therefore, Formula 5 becomes the following Formula 8.
[0047]
[Equation 8]

[0048]
As in FIG. 6, when A is 2 to the power of a and B is to the power of 2 to b, when the rotation angle is 30 degrees from Equation 8, the elliptical distance calculation block is x, α, With y, β, and √3 as inputs, subtracters 81-1 to 81-2 and 83-1, adders 83-2 and 86, multipliers 82-1 to 82-2, 85-1 to 85-2, The shifter 84-1 to 84-2 that performs bit shift operation of a-1 and b-1 are used, and the near distance e " ² Can be realized.
[0049]
If the rotation angle is 30n degrees (where n is not 3m, m and n are arbitrary integers), cos θ and sin θ in FIG. 7 are ± √3 / 2 or ± 1/2, so The distance calculation block uses selector S1 (92-1) and selector S2 (92-2) to input x, α, y, β, √3, 1, -√3, -1 as shown in FIG. Selector S1 (92-1), selector S2 (92-2), subtracters 91-1 to 91-2, 94-1, adders 94-2 and 97, multipliers 93-1 to 93-4, 96-1 to 96-2, a-1 and b-1, shifters 95-1 to 95-2 that perform bit shift operations, and a near distance e " ² Can be realized. The selector S1 (92-1) operates so as to output the value of cos θ in FIG. 9, and the selector S2 (92-2) operates so as to output the value of sin θ in FIG.
[0050]
Here, operations of the selector S1 (92-1) and the selector S2 (92-2) included in FIG. 9 will be described.
FIG. 9 shows an elliptical distance calculation block when the rotation angle is 30n degrees (where n is not equal to 3m, and m and n are arbitrary integers). The selector S1 (92-1) is shown in FIG. The selector S2 (92-2) operates to output the value of sin θ × 2 in Fig. 7. This is because bit shifts are performed by shifters 95-1 to 95-2 that perform bit shift operations of a-1 and b-1 in the subsequent stage.
[0051]
Specifically, as shown in Table 1 below, the selector S1 (92-1) and the selector S2 (92-2) have n values out of four values of −√3, −1, 1, √3. One value is selected according to the value of and output.
[0052]
[Table 1]

[0053]
Further, in FIG. 9, when the selector S ′ (102-1, 102-2) shown in FIG. 10 is used, the multipliers 93-1 to 93-4 are reduced as shown in FIG. 10, and x, α, y, With β and √3 as inputs, selector S ′ (102-1 and 102-2), subtracters 101-1 to 101-2 and 104-1, adders 104-2 and 107, and multipliers 103-1 to 103 -, 106-1 to 106-2, shifters 105-1 to 105-2 that perform bit shift operations of a-1, b-1, and a near distance e " ² Can be configured to output.
[0054]
However, the selector S ′ (102-1, 102-2) sets sig1 to sig4 shown in FIG. 10 to the following values, specifically, sig1: x−α, sig2: y−β, sig3: √3 (x -Α), sig4: A selector that operates as shown in Table 2 depending on the value of n shown in Table 2 when set to √3 (y-β).
[0055]
[Table 2]

[0056]
Further, when n = 3 m, the elliptical distance calculation block includes a selector S ′ (102-1, 102-2), multipliers 103-1 to 103-2, and a subtractor 104-1, as shown in FIG. , Shifter 112-1 to 112-2 that performs bit shift operation of subtracters 111-1 to 111-2, a, and b with x, α, y, and β as inputs , 113-1 to 113-2 multiplier and adder 114, the short distance e ″ ² Can be realized.
[0057]
Using the elliptical distance calculation block shown in FIGS. 6 to 11 described above, the specific color area detection circuit can be realized as shown in FIG. In FIG. 12, the elliptical distance calculation block 121 in FIGS. 6 to 11 calculates the recent distance e ″ from the pixel information P and the elliptical area information I. ² The multiplier 122 calculates the square r of the extraction radius r. ² And the comparator (comp) 123 calculates the near distance e " ² And square of extraction radius R ² And a determination flag F is output.
[0058]
However, the elliptical area information I in FIG. 12 is the color difference signals Cb, the enlargement ratio in the direction of Cr and the center coordinates of C1, and the block of the comparator (comp) 123 is the near distance e ″. ² And square r of extraction radius r ² And the extraction radius r ² More elliptical distance e " ² If 1 is smaller, 1 is output and the extraction radius r ² More elliptical distance e " ² This is a block that outputs 0 if is larger.
[0059]
In this embodiment, an elliptical distance calculation is performed, and the value obtained by squaring the distance is used for comparison in order to avoid the development of the square root when performing the distance comparison. However, it is not always necessary to compare using the distance. You can use a value proportional to the distance. Therefore, for example, a method such as a hexagonal approximation (see Japanese Patent Application Laid-Open No. 2002-216136) represented by the following formula 9 may be used for the distance calculation.
[0060]
[Equation 9]

[0061]
Next, the second noise reduction technique will be described.
When reducing noise contained in digital images, for example, for images of specific colors such as skin color, the filter action of the filter is strengthened to lose detail and prevent skin problems such as wrinkles and rough skin. It is possible to make it difficult to stand. In this case, there is a problem that when the process is rapidly changed at the boundary portion other than the specific color region and the specific color region, the image becomes uncomfortable.
[0062]
Therefore, by performing weighting of filter coefficients using the color information of surrounding pixels in order to change the filtering action of the filter of the target pixel, the specific color region is filtered with a coefficient corresponding to the strong filtering action. It is possible to apply a filter with a coefficient corresponding to a weak filtering action to other areas, and to gradually change the noise reduction level at the boundary between the specific color area and the area other than the specific color.
[0063]
In order to reduce noise, a digital filter having a two-dimensional tap coefficient as shown in FIG. 13A or 13B is used. In FIG. 13A or 13B, α is the filter coefficient of the pixel of interest, and when all the peripheral pixels of the pixel of interest are of a specific color, α is the minimum value, and the filter is most strongly filtered. FIG. 13 shows a 3 × 3 filter and a 5 × 5 filter, but the filter coefficient and size are not limited as long as equivalent characteristics can be obtained.
[0064]
FIG. 14 shows a characteristic diagram for deriving the filter coefficients. For each characteristic L1, L2, and L3, the vertical axis represents the filter coefficient α of the target pixel, and the horizontal axis represents the number of specific colors of the surrounding pixels.
[0065]
In FIG. 14, when all the peripheral pixels are not the specific color, the filter coefficient α shown on the vertical axis is F_max, and when all the peripheral pixels are the specific color, the filter coefficient is F_min.
[0066]
As shown on the horizontal axis in FIG. 14, when the number of specific colors of the surrounding pixels is counted, the filter coefficient α of the target pixel shown on the vertical axis depends on the count value shown on the horizontal axis. It can be expressed as the following formula 10.
[0067]
[Expression 10]

[0068]
Here, if the filter coefficient calculation block for obtaining the filter coefficient α of the target pixel from the count value of the number of specific colors of the peripheral pixels shown on the horizontal axis is an f_nr (count) block 152, the noise reduction block is shown in FIG. As shown in FIG. 15, the specific color determination flag F, the luminance signal Y, and the color difference signals Cr and Cb for determining the detection of the specific color area such as the skin color output from the specific color area detection circuit shown in FIG. F_nr () of the filter coefficient calculation block for obtaining the filter coefficient α of the target pixel from the count value of the number of specific colors of the peripheral pixels based on the delay line (DL) 151 and the specific color determination flag F count) block 152, a two-dimensional filter for filtering the luminance signal y by a two-dimensional tap coefficient of the filter coefficient α of the target pixel. A digital filter 153 can be realized.
[0069]
For example, when the characteristic L2 in FIG. 14 is taken as an example, the above-described formula 10 can be expressed as the following formula 11.
[0070]
[Expression 11]

[0071]
In Equation 11, Filter_size of the denominator of the two items is the number of elements of the filter, and count is the number of specific colors of the peripheral pixels. When Equation 11 is realized by hardware, the noise reduction block, as shown in FIG. 16, is a specific color determination flag F for determining the detection of a specific color region such as skin color output from the specific color region detection circuit shown in FIG. The luminance signal Y and the color difference signals Cr and Cb are added to the delay line (DL) 161 that simultaneously synchronizes the luminance signal Y and the color difference signals Cr and Cb, and the specific color number of the peripheral pixels is added based on the specific color determination flag F. an adder 164 that outputs x, a subtractor 162 that outputs F_min by subtracting F_min from F_max, a divider 163 that divides the subtracted value by Filter_size and outputs a divided value, and a specific color number x of surrounding pixels A multiplier 165 that multiplies the value and outputs the multiplication value, an adder 166 that adds the multiplication value and F_max to output the filter coefficient α of the pixel of interest, and a filter unit of the pixel of interest Constitute a two-dimensional digital filter 167 to perform the filtering action on the luminance signal y by 2-dimensional tap coefficients alpha, it becomes possible to achieve. However, since the noise reduction block of FIG. 16 uses the divider 163, the hardware scale becomes large.
[0072]
Thus, for example, consider a case where a value of F_max is 10 and a value of F_min is 1 using a 3 × 3 filter.
At this time, Equation 11 becomes the following Equation 12.
[0073]
[Expression 12]

[0074]
When the filter coefficient α is expressed by Equation 12, the noise reduction block, as shown in FIG. 17, determines whether to detect a specific color area such as a flesh color output from the specific color area detection circuit shown in FIG. A peripheral line is obtained by adding the number of specific colors of peripheral pixels based on a delay line (DL) 171 and a specific color determination flag F that simultaneously synchronizes the color determination flag F, the luminance signal Y, and the color difference signals Cr and Cb. An adder 172 that outputs the specific color number x, a subtractor 173 that subtracts F_max from the specific color number x of the surrounding pixels and outputs a filter coefficient α of the pixel of interest, and a two-dimensional tap coefficient of the filter coefficient α of the pixel of interest Can be realized by a two-dimensional digital filter 174 that filters the luminance signal y, thereby eliminating the divider 163 and allowing a simple circuit configuration. It is.
[0075]
In order to simplify the circuit configuration as an example, Equation 12 has been described. However, the equation for obtaining the filter coefficient α is a general equation represented by Equation 10 such as the characteristics L1 to L3 in FIG. It doesn't matter.
For example, the following equation 13 may be used.
[0076]
[Formula 13]

[0077]
However, αmax is a value of α when all the peripheral pixels are not of a specific color.
[0078]
Further, an aperture correction technique in which the gain changes smoothly at the boundary portion between the third specific color region and the region other than the specific color will be described.
In digital image signal processing, aperture correction (f-specific correction (frequency characteristic correction)) is often performed to correct the image resolution and resolution and improve the appearance.
[0079]
For example, as shown in FIG. 18, the luminance signal Y and the color difference signals Cr and Cb of the input signal are delayed by the delay line (DL) 181 and the luminance is obtained by the high pass filter (HPF) 182 with respect to the main image line 184 that is synchronized in the vertical direction. A method is known in which the high frequency component 185 of the image is extracted from the signal Y and the color difference signals Cr and Cb, and is added again to the luminance signal Y and the color difference signals Cr and Cb of the image main line 184 by the adder 183.
[0080]
However, in a specific color region such as a skin color, it is necessary to reduce the aperture correction signal in order to make the skin region troubles such as wrinkles and rough skin less noticeable by losing detail.
[0081]
However, if the aperture correction signal is reduced for the entire area, the image will be blurred, so only the aperture correction signal for the specific color area is reduced, and the aperture correction signal for the area other than the specific color is not changed. It is desirable. Further, in order not to give a sense of incongruity to the output image, the aperture correction signal needs to change smoothly in the specific color region and the region other than the specific color.
[0082]
Therefore, as shown in FIG. 19, the luminance signal Y and the color difference signals Cr and Cb of the input signal are synchronized in the vertical direction by the delay line (DL) 191, and the luminance signal Y is vertical by the vertical low-pass filter (V-LPF) 192. The low frequency component in the direction is extracted, the predetermined band frequency component in the horizontal direction of the luminance signal Y is extracted by the horizontal bandpass filter (H-BPF) 194, and the vertical of the luminance signal Y is extracted by the vertical bandpass filter (V-BPF) 193 A predetermined band frequency component in the direction is extracted, a horizontal low-pass frequency component of the luminance signal Y is extracted by the horizontal low-pass filter (H-LPF) 195, an aperture correction signal is output by the adder 197, and the specific color determination flag F Based on the color information around the target pixel, the gain calculation block 196 counts the number of specific colors of the peripheral pixels (c The gain of the pixel of interest is obtained from the value of (out), the gain is given to the aperture correction signal, and the multiplier 198 multiplies the aperture correction signal by the gain.
[0083]
When the number of specific colors of the peripheral pixels is counted, the gain of the aperture correction signal of the pixel of interest depends on the value of the count, and can be expressed as Equation 14.
[0084]
[Expression 14]

[0085]
A gain calculation block 196 shown in FIG. 19 is obtained by realizing the equation (14) by hardware.
[0086]
Here, when the number of specific colors of the peripheral pixels on the horizontal axis is counted as indicated by the characteristics L1, L2, and L3 shown in the gain derivation diagram of the aperture correction signal in FIG. By smoothly changing the aperture correction signal only in the specific color region, the aperture correction signal can be changed smoothly at the boundary with the region other than the specific color.
[0087]
Here, taking the characteristic L2 of FIG. 20 as an example, the equation for calculating the gain of the aperture correction signal is as shown in the following equation (15).
[0088]
[Expression 15]

[0089]
When Expression 15 is realized by hardware, the gain calculation block 196 shown in FIG. 19 adds the number of specific colors of the peripheral pixels based on the specific color determination flag F as shown in FIG. An adder 211 that outputs the number x, a subtractor 212 that subtracts G_min from G_max and outputs a subtraction value, a divider 213 that divides the subtraction value by Filter_size and outputs a division value, and a specific color number x of surrounding pixels A multiplier 214 that multiplies the division value and outputs the multiplication value, and an adder 215 that adds the multiplication value and G_max to output the gain G of the aperture correction signal of the target pixel can be realized. is there. In FIG. 21, the divider 213 is required. Here, if X_value in FIG. 20 is 2 to the x ′ power, Equation 15 can be transformed into Equation 16.
[0090]
[Expression 16]

[0091]
Since X_value is 2 to the power x ', the division in Equation 16 can be realized by a shifter that performs bit shift of x'. FIG. 22 shows a gain calculation block in the case where the division in equation (16) is realized by a shifter that performs bit shift of x ′. Based on the specific color determination flag F, the number of specific colors of the peripheral pixels is added to the surroundings. An adder 221 that outputs the specific color number x of the pixel, a shifter 224 that performs a bit shift of x ′ of G_max and outputs a bit shift value, and multiplies the specific color number x of the surrounding pixels by the bit shift value of x ′ of G_max. Thus, it can be realized by a multiplier 222 that outputs a multiplication value and an adder 223 that adds the multiplication value and G_max and outputs the gain G of the aperture correction signal of the target pixel.
[0092]
Here, the expression 16 is described as an example for deriving the gain of the aperture correction signal. However, it may be a general expression represented by the expression 14 like the characteristics L1 to L3 in FIG.
For example, the following equation 17 may be used.
[0093]
[Expression 17]

[0094]
However, gmax in Expression 17 is a value of g (gain) when all the peripheral pixels are not of a specific color.
[0095]
As described above, the aperture correction block synchronizes the input luminance signal Y and the color difference signals Cr and Cb in the vertical direction with the delay line (DL) 231 as shown in FIG. 23 and the vertical low-pass filter (V-LPF) 232. A low frequency component in the vertical direction of the luminance signal Y is extracted, a predetermined band frequency component in the horizontal direction of the luminance signal Y is extracted by a horizontal band pass filter (H-BPF) 234, and a vertical band pass filter (V-BPF) 233 is extracted. Extracts the predetermined band frequency component of the luminance signal Y in the vertical direction, extracts the low frequency component in the horizontal direction of the luminance signal Y by the horizontal low-pass filter (H-LPF) 235, outputs the aperture correction signal by the adder 236, and specifies it. Based on the color information around the pixel of interest of the color determination flag F, the gain calculation block 237 The gain of the pixel of interest is obtained from the count value of the number of specific colors, a gain is given to the aperture correction signal, the gain is multiplied by the aperture correction signal by the multiplier 238, and this multiplication value is added to the luminance signal by the adder 239. y is added to the color difference signals Cr and Cb.
[0096]
“V-LPF” 232 in FIG. 23 is a low-pass filter in the vertical direction, and “V-BPF” 233 is a band-pass filter in the vertical direction. Similarly, “H-LPF” 235 and “H-BPF” 234 indicate a horizontal low-pass filter and a horizontal band-pass filter, respectively. Although not shown in the figure, a block for processing that the minute amplitude is regarded as noise as a coring unit and that the aperture signal of the portion regarded as noise is set to 0 may be added after the multiplier 238. .
[0097]
Fourthly, the case where the output of the determination flag F of the comparator (comp) 123 portion is a multi-bit signal corresponding to the distance in the specific color area detection in the specific color area detection circuit shown in FIG.
[0098]
In the specific color area detection, by expressing the determination result of whether or not it is included in the specific color area with multiple bits, it is possible to indicate “specific color likelihood” as shown in FIG. The closer a certain pixel is to the center of the specific color area, the closer to the specific color.
[0099]
For example, the case where it is assumed in FIG. 24 that the specific color likelihood of the determination flag F is expressed by 2 bits will be described. The weight is set to 3 when it is included in the area closest to the center, 2 is set when it is included in the area closest to the center, and the case where it is included in the outermost area in the elliptical area in FIG. 1 and 0 if not included in the elliptical region. By assigning a larger value to a certain pixel closer to the center of the specific color area in this way, it becomes possible to determine how close to the center of the elliptical area using the magnitude of the value of the determination result. .
[0100]
As an example, the case where the specific color-likeness is expressed by 2 bits has been described, but any number of bits may be used to express the specific color-likeness.
[0101]
When the decision result of the specific color area detection block is expressed in multiple bits, the filter coefficient of the pixel of interest changes more smoothly in the noise reduction unit than in the case of expressing in 1 bit, and the aperture correction signal gain is smoother in the aperture correction unit. Will be switched to. Therefore, the specific color determination flag F shown in FIG. 12, FIG. 15 or the like is multi-bit, but the specific color determination result is expressed by 1 bit even if the configuration of other blocks is not changed. A smoother process can be realized at the boundary between the specific color region and the region other than the specific color.
[0102]
According to the present embodiment described above, noise of a specific color such as skin color can be reduced. For example, noise reduction (NR) is intentionally applied to an area where there is almost no noise to weaken the aperture correction signal. By losing the details, for example, in the skin color region, it becomes possible to make skin troubles such as “acne” and “stain” less noticeable and to reproduce beautiful skin. In addition, other color regions can be prevented from changing.
[0103]
Needless to say, the present invention is not limited to the above-described embodiment, and other configurations can be appropriately employed without departing from the scope of the claims of the present invention.
[0104]
【The invention's effect】
The signal processing apparatus according to the present invention provides image data picked up by the image pickup means. of Based on luminance and color difference signals The Color difference value of each pixel in the color difference plane The result of determining whether or not the image is included in the specific color area is expressed by a multi-bit signal having a larger value as it is closer to the center of the specific color area Specific color area detection means and specific color area detection means Based on the determination result, the filter coefficient of the filter that reduces the noise of the pixel of interest is decreased as the count value of the multi-bit signal for the surrounding pixels is increased. Since noise reduction means is provided and signal processing is performed in a specific color area, a specific color area such as skin color or sky blue is used when performing image processing. By expressing the determination result of whether or not to be included in multiple bits, the filter coefficient of the pixel of interest changes smoothly compared to the case where it is expressed by 1 bit. This The process is smoothly changed at the border between the specific color area and the specific color area to reduce noise so that the image quality does not feel strange. So you can For example, noise of a specific color such as skin color can be reduced to generate a smooth image, and the image quality can be improved.
[0105]
In the signal processing device according to the present invention, the specific color area detecting means is ,color In the difference plane Move and rotate the distance of each pixel to the elliptical specific color area and then the square value of the distance compressed and expanded in the vertical and horizontal directions, and move and rotate the elliptical specific color area on the color difference plane and move the vertical and horizontal directions Since the judgment is made by comparing with the square of the radius of the compressed / expanded circle, the specific color area detection means can be realized with a simple circuit. There is an effect that can be done.
[0106]
Moreover, the signal processing apparatus of the present invention is as described above. The specific color area detection means performs the calculation by limiting the compression / expansion ratio to a power of 2 or limiting the rotation angle to 30 degrees, 60 degrees, 120 degrees, or 150 degrees. Therefore, the specific color area detection means is realized with a simpler circuit. There is an effect that can be done.
[0107]
The signal processing apparatus according to the present invention also provides image data picked up by the image pickup means. of Based on luminance and color difference signals The Color difference value of each pixel in the color difference plane The result of determining whether or not the image is included in the specific color area is expressed by a multi-bit signal having a larger value as it is closer to the center of the specific color area Specific color area detection means and specific color area detection means Based on the determination result, the gain of the aperture correction signal for the target pixel is decreased as the count value of the multi-bit signal for the peripheral pixel is increased. Aperture correction means is provided to perform signal processing in a specific color area, so when performing image processing, a specific color area such as skin color or sky blue By expressing the determination result of whether or not to be included in multiple bits, the gain of the aperture correction signal is smoothly switched compared to the case of expressing with 1 bit. This Smoothly change the process at the boundary between the specific color area and the specific color area, and perform aperture correction so as not to make the image quality uncomfortable. So you can For example, a specific color such as skin color Aperture correction There is an effect that it is possible to generate a smooth image and improve the image quality.
[0108]
In the signal processing device according to the present invention, the specific color area detecting means is ,color In the difference plane Move and rotate the distance of each pixel to the elliptical specific color area and then the square value of the distance compressed and expanded in the vertical and horizontal directions, and move and rotate the elliptical specific color area on the color difference plane and move the vertical and horizontal directions Since the judgment is made by comparing with the square of the radius of the compressed / expanded circle, the specific color area detection means can be realized with a simple circuit. There is an effect that can be done.
[0109]
Moreover, the signal processing apparatus of the present invention is as described above. The specific color area detection means performs the calculation by limiting the compression / expansion ratio to a power of 2 or limiting the rotation angle to 30 degrees, 60 degrees, 120 degrees, or 150 degrees. Therefore, the specific color area detection means is realized with a simpler circuit. There is an effect that can be done.
[0110]
Further, the signal processing method of the present invention provides image data captured by the imaging means. of Based on luminance and color difference signals The Color difference value of each pixel in the color difference plane The result of determining whether or not the image is included in the specific color area is expressed by a multi-bit signal having a larger value as it is closer to the center of the specific color area Specific color area detection step and specific color area detection step Based on the determination result, the filter coefficient of the filter that reduces the noise of the target pixel is decreased as the count value of the multi-bit signal for the peripheral pixel is increased. Noise reduction step and signal processing in a specific color area, so when performing image processing a specific color area such as skin color or sky blue By expressing the determination result of whether or not to be included in multiple bits, the filter coefficient of the pixel of interest changes smoothly compared to the case where it is expressed by 1 bit. This The process is smoothly changed at the border between the specific color area and the specific color area to reduce noise so that the image quality does not feel strange. So you can For example, noise of a specific color such as skin color can be reduced to generate a smooth image, and the image quality can be improved.
[0111]
In the signal processing method of the present invention, the specific color region detecting step is ,color In the difference plane Move and rotate the distance of each pixel to the elliptical specific color area and then the square value of the distance compressed and expanded in the vertical and horizontal directions, and move and rotate the elliptical specific color area on the color difference plane and move the vertical and horizontal directions Since the judgment is made by comparing with the square value of the radius of the compressed and expanded circle, the block that performs the specific color area detection step is realized with a simple circuit. There is an effect that can be done.
[0112]
Further, the signal processing method of the present invention is as described above. The specific color area step is calculated by limiting the compression / expansion ratio to a power of 2 and / or limiting the rotation angle to 30 degrees, 60 degrees, 120 degrees, and 150 degrees. , Block that performs specific color area detection step is realized with a simpler circuit There is an effect that can be done.
[0113]
Further, the signal processing method of the present invention provides image data captured by the imaging means. of Based on luminance and color difference signals The Color difference value of each pixel in the color difference plane The result of determining whether or not the image is included in the specific color area is expressed by a multi-bit signal having a larger value as it is closer to the center of the specific color area Specific color area detection step and specific color area detection step Based on the determination result, the gain of the aperture correction signal for the target pixel is decreased as the count value of the multi-bit signal for the peripheral pixel is increased. Since the aperture correction step is included and signal processing is performed in the specific color area, the specific color area such as skin color or sky blue is used when performing image processing. By expressing the determination result of whether or not to be included in multiple bits, the gain of the aperture correction signal is smoothly switched compared to the case of expressing with 1 bit. This Smoothly change the process at the boundary between the specific color area and the specific color area, and perform aperture correction so as not to make the image quality uncomfortable. So you can For example, a specific color such as skin color Aperture correction There is an effect that it is possible to generate a smooth image and improve the image quality.
[0114]
In the signal processing method of the present invention, the specific color region detecting step is ,color In the difference plane Move and rotate the distance of each pixel to the elliptical specific color area and then the square value of the distance compressed and expanded in the vertical and horizontal directions, and move and rotate the elliptical specific color area on the color difference plane and move the vertical and horizontal directions Since the judgment is made by comparing with the square value of the radius of the compressed and expanded circle, the block that performs the specific color area detection step is realized with a simple circuit. There is an effect that can be done.
[0115]
Further, the signal processing method of the present invention is as described above. The specific color area step is calculated by limiting the compression / expansion ratio to a power of 2 and / or limiting the rotation angle to 30 degrees, 60 degrees, 120 degrees, and 150 degrees. , Block that performs specific color area detection step is realized with a simpler circuit There is an effect that can be done.
[Brief description of the drawings]
FIG. 1 is a block diagram of a digital still camera.
FIG. 2 is a block diagram of a camera system LSI.
FIG. 3 is a block diagram of a camera signal processing unit applied to the embodiment of the present invention.
FIG. 4 is a diagram illustrating a color difference distribution of a specific color (skin color).
FIG. 5 is a diagram showing an elliptical distance calculation procedure;
FIG. 6 is a diagram showing an elliptical distance calculation block.
FIG. 7 is a diagram showing an elliptical distance calculation block in which expansion and compression are realized by shift.
FIG. 8 is a diagram showing an elliptical distance calculation block when the rotation angle is 30 degrees.
FIG. 9 is a diagram showing an elliptical distance calculation block when the rotation angle is 30 n degrees (where n ≠ 3m: m, n is an arbitrary integer).
FIG. 10 is a diagram showing an elliptical distance calculation block using a selector S ′.
FIG. 11 is a diagram showing an elliptical distance calculation block when n = 3 m.
FIG. 12 is a diagram illustrating a specific color area detection circuit.
13 is a diagram illustrating an NR filter, FIG. 13A is a 3 × 3 filter, and FIG. 13B is a 5 × 5 filter.
FIG. 14 is a diagram showing derivation of a filter coefficient α.
FIG. 15 is a diagram illustrating a noise reduction block (1).
FIG. 16 is a diagram illustrating a noise reduction block (2).
FIG. 17 is a diagram illustrating a noise reduction block (3).
FIG. 18 is a diagram illustrating general aperture correction.
FIG. 19 is a diagram illustrating derivation of an aperture correction signal.
FIG. 20 is a diagram illustrating derivation of the gain of the aperture correction signal.
FIG. 21 is a diagram showing a gain calculation block (1).
FIG. 22 is a diagram showing a gain calculation block (2).
FIG. 23 is a diagram illustrating an aperture correction block.
FIG. 24 is a diagram illustrating the likelihood of a specific color.
[Explanation of symbols]
DESCRIPTION OF SYMBOLS 1 ... Lens, 2 ... Aperture, 3 ... Shutter, 4 ... Image sensor, 5 ... TG (timing generator), 6 ... Front end, 7 ... Camera system LSI, 8 ... Image memory, 9 ...... Microcomputer, 10 ... External storage medium, 11 ... Image monitor, 21 ... Image detection unit, 22 ... Camera signal processing unit, 23 ... Memory I / F, 24 ... Memory controller, 25 ... Microcomputer I / F, 26 ... Image compression / decompression unit, 27 ... Monitor I / F, 31 ... Specific color area detection unit, 32 ... DL (delay line), 33 ... NR (noise reduction) unit, 34 ...... AP (aperture) correction unit, 41..elliptical area (skin color), 42..elliptical area (light blue), 51 .... moving, 52, 53 ... rotation, 54 ... enlargement / compression ...... Oval distance calculation Lock

Claims (12)

Based-out the luminance signal and color difference signals of image data taken by the imaging means, whether or not the determination result color difference values of each pixel in the color difference plane is included in a specific color area, close to the center of the specific color region A specific color area detecting means for expressing a multi-bit signal having a large value ;
Noise reduction means for reducing the filter coefficient of the filter for reducing the noise of the pixel of interest based on the determination result of the specific color area detection means as the count value of the multi-bit signal for the surrounding pixels increases signal processing apparatus that performs signal processing in the color area. The signal processing device according to claim 1,
The specific color area detecting means includes a square value of a distance obtained by compressing and expanding the length and width after moving and rotating the distance of each pixel with respect to the elliptical specific color area on the color difference plane, and the elliptical shape on the color difference plane. A signal processing device that makes a determination by comparing a square value of a radius of a circle that has been compressed and expanded in length and breadth after moving and rotating the specific color region . The signal processing device according to claim 2 ,
The specific color area detecting means limits at least one of the compression / expansion ratio to a power of 2 and the rotation angle to 30 degrees, 60 degrees, 120 degrees, and 150 degrees. A signal processing device that performs calculations . Based-out the luminance signal and color difference signals of image data taken by the imaging means, whether or not the determination result color difference values of each pixel in the color difference plane is included in a specific color area, close to the center of the specific color region A specific color area detecting means for expressing a multi-bit signal having a large value ;
Aperture correction means for reducing the gain of the aperture correction signal for the target pixel based on the determination result of the specific color area detection means as the count value of the multi-bit signal for the peripheral pixel is increased, and in the specific color area signal processing apparatus that performs signal processing. The signal processing device according to claim 4,
The specific color area detecting means includes a square value of a distance obtained by compressing and expanding the length and width after moving and rotating the distance of each pixel with respect to the elliptical specific color area on the color difference plane, and the elliptical shape on the color difference plane. A signal processing device that makes a determination by comparing a square value of a radius of a circle that has been compressed and expanded in length and breadth after moving and rotating the specific color region . The signal processing device according to claim 5 ,
The specific color area detecting means limits at least one of the compression / expansion ratio to a power of 2 and the rotation angle to 30 degrees, 60 degrees, 120 degrees, and 150 degrees. A signal processing device that performs calculations . Based-out the luminance signal and color difference signals of image data taken by the imaging means, whether or not the determination result color difference values of each pixel in the color difference plane is included in a specific color area, close to the center of the specific color region A specific color area detection step represented by a multi-bit signal having a large value ;
A noise reduction step of reducing a filter coefficient of a filter that reduces noise of a pixel of interest based on a determination result of the specific color region detection step , as the count value of the multi-bit signal for the peripheral pixel is larger signal processing method to perform signal processing in the color area. The signal processing method according to claim 7, wherein
The specific color area step includes the square value of the distance obtained by compressing and expanding the vertical and horizontal directions after moving and rotating the distance of each pixel with respect to the elliptical specific color area on the color difference plane, and the elliptical shape on the color difference plane. A signal processing method for making a determination by comparing a square value of a radius of a circle obtained by moving and rotating a specific color area and then compressing and enlarging the vertical and horizontal directions . The signal processing device according to claim 8 , wherein
The specific color region detection step may limit at least one of the compression / expansion ratio to a power of 2 and the rotation angle to 30 degrees, 60 degrees, 120 degrees, and 150 degrees. A signal processing method for performing calculations . Based-out the luminance signal and color difference signals of image data taken by the imaging means, whether or not the determination result color difference values of each pixel in the color difference plane is included in a specific color area, close to the center of the specific color region A specific color area detection step represented by a multi-bit signal having a large value ;
An aperture correction step for reducing the gain of the aperture correction signal for the pixel of interest based on the determination result of the specific color region detection step as the count value of the multi-bit signal for the peripheral pixel increases . signal processing method to perform signal processing. The signal processing method according to claim 10.
The specific color area step includes the square value of the distance obtained by compressing and expanding the vertical and horizontal directions after moving and rotating the distance of each pixel with respect to the elliptical specific color area on the color difference plane, and the elliptical shape on the color difference plane. A signal processing method for making a determination by comparing a square value of a radius of a circle obtained by moving and rotating a specific color area and then compressing and enlarging the vertical and horizontal directions . The signal processing device according to claim 11 ,
The specific color region detection step may limit at least one of the compression / expansion ratio to a power of 2 and the rotation angle to 30 degrees, 60 degrees, 120 degrees, and 150 degrees. A signal processing method for performing calculations .

Images