JP3724525B2 - Image processing method and image processing apparatus - Google Patents

Description

ここで、Ｖ_i ＝（Ｌ_i,Ｃ_i,Ｈ_i,ｘ_i,ｙ_i ）、Ｖ_j ＝（Ｌ_j,Ｃ_j,Ｈ_j,ｘ_j,ｙ_j ）である。
【００４４】
また、複数の画素から構成されるクラスタのクラスタ中心には、クラスタに属する画素の明度、彩度、色相と画像平面図におけるＸ座標、Ｙ座標からなる５次元ベクトルの平均ベクトルを用いた。例えば、クラスタ中心Ｖ_i （１≦ｉ≦Ｎｃ）は（２）式を用いることで算出することができる。
【数２】

ここで、Ｎｃ_i はクラスタｃ_i に属する画素の総数であり、Ｌ_ik，Ｃ_ik，Ｈ_ik，ｘ_ik，ｙ_ikは、それぞれ、クラスタｃ_i に属するｋ番目の画素の明度、彩度、色相、Ｘ座標、Ｙ座標である。
【００４５】
Ｓ１０３では、Ｓ１０１で抽出されたクラスタに属する画素について、どのクラスタ中心との距離が最も近いかを調べて、クラスタを再構成する。ここでは、画素の明度、彩度、色相と画像平面のＸ座標とＹ座標の５次元からなる特徴空間において、再配置の対象の画素とすべてのクラスタ中心との重み付き距離を算出する。ここで、処理の効率化のため、すべてのクラスタではなく再配置の対象の画素の近傍に存在するクラスタ中心についてのみの重み付き距離を算出してもよい。この重み付き距離が最も小さいクラスタ中心に対応するクラスタに、対象の画素が属するように画素を再配置する。重み付き距離ｄ_i は、例えば、対象の画素とクラスタ中心Ｖ_i （１≦ｉ≦Ｎｃ）とから（３）式によって求めることができる。
【数３】

ここで、（Ｌ_T,Ｃ_T,Ｈ_T,ｘ_T,ｙ_T ）は、対象画素Ｔの前述した５次元特徴空間上の座標であり、Ｖ_i ＝（Ｌ_i,Ｃ_i,Ｈ_i,ｘ_i,ｙ_i ）である。また、ａ，ｂ，ｃ，ｄ，ｅはそれぞれの特徴に対応する重み係数である。この重み係数を適当に定めることで、生成されるグループを調節することができる。例えば、ｃを比較的大きくして、ｄとｅを比較的小さめに設定すれば、色相が近ければ画像平面上における距離が少々遠くても１つのクラスタとして構成することができる。
【００４６】
上述の重み付き距離の計算の結果、ある画素とクラスタ中心との最も小さい重み付き距離があらかじめ設定された閾値Ｔｈ₄ よりも大きい場合には、より適切なクラスタを作成するために、その画素１つからなるクラスタを新たに生成することもできる。
【００４７】
Ｓ１０４では、Ｓ１０３で、再構成されたクラスタのクラスタ中心を算出する。このクラスタ中心の算出方法は、Ｓ１０２で述べた方法と同様である。
【００４８】
Ｓ１０５では、Ｓ１０３で行なわれた画素の再配置でクラスタが変更されたか否かを調べて、変更されていなければ、画素の再配置は収束したものと見なして、クラスタ抽出処理を終了する。変更されている場合には、Ｓ１０３に戻って処理を繰り返す。
【００４９】
ここで説明した例では、重要色と重要色に類似する色に限定した形でクラスタを抽出したが、重要色との距離により色を複数のクラスに分類し、代表色のクラスによって再配置に使用するパラメータや閾値を変更するように構成することもできる。例えば、重要色と、重要色に類似する色と、その他の色の３つのクラスを定義して、重要色に近いクラスほど小さなクラスタが得られるように閾値を設定しておくことで、重要な色ほど詳細な色分布が得られて、結果的により再現性の高い色変換が実現できる。
【００５０】
図２に戻り、クラスタ解析部１６は、クラスタ抽出部１５で抽出されたクラスタ内の色分布を認識し、必要に応じてこの色分布によるクラスタの分割、統合および削除を行なう。この例では、クラスタに属する画素の色が分散しているか集中しているかにより、それぞれ、グラデーションとフラットの２つの色分布に分類する。そして、類似する色分布を持つクラスタを統合した後に微小クラスタを削除する。
【００５１】
図５は、本発明の第１の実施の形態を実現する構成例におけるクラスタ解析部１６のクラスタの色分布認識処理の動作の一例を示すフローチャートである。Ｓ１１１では、色分布が認識されていないクラスタｃ_i を１つ選択する。以降の処理は、このクラスタｃ_i を対象に行なう。
【００５２】
Ｓ１１２では、クラスタ内の色分布を判断するためにクラスタに属する画素のＬ^* Ｃ^* Ｈ°色空間における分散ベクトルを算出する。例えば、（４）式を用いることによりクラスタｃ_i の分散ベクトルＳ_i ²（１≦ｉ≦Ｎｃ）を算出することができる。
【数４】

ここで、クラスタｃ_i のクラスタ中心Ｖ_i ＝（Ｌ_i,Ｃ_i,Ｈ_i,ｘ_i,ｙ_i ）である。以降、Ｍ_i ＝（Ｌ_i,Ｃ_i,Ｈ_i ）をクラスタｃ_i の代表色と称する。
【００５３】
Ｓ１１３では、クラスタに属する画素のＬ^* Ｃ^* Ｈ°色空間における分散ベクトルの大きさがあらかじめ設定された閾値Ｔｈ₅ よりも小さいか否かを調べて、小さければクラスタ内の色分布はほぼ一様であると判断してＳ１１４にてクラスタｃ_i に対して「フラット」という属性を付加する。大きい場合にはクラスタ内の色分布はある幅で変化していると判断してＳ１１５にてクラスタｃ_i に対して「グラデーション」という属性を付加する。もちろん、明度、彩度、色相の各分散値と閾値Ｔｈ₅'とを比較してもよいし、明度、彩度、色相の分散値ごとに異なる閾値Ｔｈ_5L，Ｔｈ_5C，Ｔｈ_5H を設けて比較してもよい。
【００５４】
Ｓ１１６では、まだ色分布を認識していないクラスタが残っているか否かを調べて、残っていれば、Ｓ１１１に戻って処理を繰り返す。残っていなければ、色分布の認識処理を終了する。
【００５５】
ここでは、クラスタの色分布を調べるために分散ベクトルのみを用いたが、本発明はこれに限定するものではなく、色分布がほぼ均一である「フラット」と滑らかに色が変化している「グラデーション」の２つの色分布に分類できる方法であればどんな方法を用いてもよい。また、例えば上述の処理に続いて、分散ベクトルの大きさがあらかじめ設定された閾値Ｔｈ₅ よりも大きい場合には、さらにクラスタ内の隣接する画素の色の変化率を調べることにより、色の変化率があらかじめ設定している閾値Ｔｈ₅"よりも小さい場合は「グラデーション」、大きな場合には「その他」という属性を付加するように構成することもできる。
【００５６】
図６は、本発明の第１の実施の形態を実現する構成例におけるクラスタ解析部１６のクラスタの統合と削除処理の動作の一例を示すフローチャートである。まず、Ｓ１２１では、統合および削除の処理が行なわれていないクラスタｃ_i を１つ選択する。Ｓ１２２では、クラスタｃ_i との比較が行なわれていないクラスタｃ_j を１つ選択する。
【００５７】
Ｓ１２３では、クラスタｃ_i とクラスタｃ_j の代表色Ｍ_i とＭ_j との差のベクトルの大きさがあらかじめ設定された閾値Ｔｈ₆ よりも小さいか否かを調べて、小さければさらにＳ１２４で、クラスタｃ_i とクラスタｃ_j の分散ベクトルＳ_i ²とＳ_j ²との差のベクトルの大きさがあらかじめ設定された閾値Ｔｈ₇ よりも小さいか否かを調べる。これも小さければ、２つのクラスタｃ_i とクラスタｃ_j の色分布は類似していると判断して、Ｓ１２５において、クラスタｃ_i にクラスタｃ_j を統合する。統合した際には、クラスタｃ_j は削除して、統合により更新されたクラスタｃ_i は再度クラスタの代表色Ｍ_i および分散ベクトルＳ_i ²を算出しておく。Ｓ１２３およびＳ１２４における比較時に条件を満たさなかった場合には、Ｓ１２６の処理へ進む。
【００５８】
Ｓ１２６では、クラスタｃ_i と比較する対象となるクラスタｃ_j がまだ残っているか否かを調べて、残っていればＳ１２２に戻って、処理を繰り返す。残っていなければ、Ｓ１２７に進む。
【００５９】
Ｓ１２７では、処理対象となるクラスタｃ_i に属する画素の総数Ｎｃ_i があらかじめ設定され閾値Ｔｈ₈ よりも小さいか否かを調べる。そして、小さければ、クラスタｃ_i は孤立した微小クラスタであると判断して、Ｓ１２８において削除する。大きい場合はそのままＳ１２９の処理へ進む。
【００６０】
Ｓ１２９では、Ｓ１２１で選択されていないクラスタｃ_i が残っているか否かを調べて、残っていればＳ１２１に戻って、処理を繰り返す。残っていなければ、クラスタの統合および削除の処理を終了する。
【００６１】
この例では、クラスタの統合条件として、２つのクラスタの代表色と色分布が類似していることを用いたが、大きなレンジを持つグラデーションの場合に複数にクラスタに分割されてしまうという不都合が生じる。したがって、より高精度な色変換装置を構成する場合には、グラデーションの色分布を持つクラスタの隣接部の色の変化率を調べて、これが小さい場合に２つのクラスタを統合するように構成すればよい。
【００６２】
図２のクラスタ別色変換部１７は、クラスタ解析部１６で認識されたクラスタの色分布に基づいて、クラスタごとに色変換用のプロファイルを作成する。そして、各クラスタに属する画素集合に対して対応するプロファイルを適用して色変換を行なう。この際、すべてのクラスタごとに異なるプロファイルを作成する必要はなく、同一のプロファイルを用いても不都合が生じなければ、複数のクラスタに対して１つのプロファイルを使用して色変換を行なうこともできる。ここで、プロファイルとは、ある色をある色に変換する変換ベクトルの集合であり、プロファイルを画像に対して適用することで、プロファイルに記述してある変換ベクトルに基づいて色変換を行なうことができる。この際、一般的にはすべての色に対する変換ベクトルを定義することは困難なため、通常、変換される色に対応する変換ベクトルが存在しない場合には、近傍の変換ベクトルを複数用いることで色変換を行なう。
【００６３】
図７は、本発明の第１の実施の形態を実現する構成例においてクラスタ別色変換部１７が行なう色変換とクラスタの色分布との対応の一例の説明図である。図中、一般クラスタとは、クラスタ抽出部１５およびクラスタ解析部１６により抽出されたクラスタを指し、特殊クラスタとは、画像データの画素の中でどのクラスタにも属さない画素の集合を指す。
【００６４】
ここでは、一般クラスタについて、入力画像のフラット部分については色識別性を保つことができるように、色再現範囲内への色変換を行なう。基本的には、色再現範囲内であれば色変換せず、色再現範囲外の色を含む場合には、予め設定されている初期プロファイルを用いる。また、グラデーション部分については階調性を保つことができるように色再現範囲内への色変換を行なう。色再現範囲内の場合には色変換せず、色再現範囲外の色を含む場合には、クラスタ内の色分布に依存したプロファイルを作成して色変換を行なう。特殊クラスタについては、予め設定されている初期プロファイルを用いて色変換を行なう。
【００６５】
図８は、本発明の第１の実施の形態を実現する構成例におけるクラスタ別色変換部１７の動作の一例を示すフローチャートである。まず、Ｓ１３１では、未処理のクラスタｃ_i を１つ選択する。以降の処理はここで選択されたクラスタｃ_i を対象のクラスタとする。
【００６６】
Ｓ１３２では、対象のクラスタｃ_i の色分布がフラットであるか否かを調べる。これは、クラスタ解析部１６で求めた色分布により判断する。対象のクラスタｃ_i の色分布がフラットである場合には、Ｓ１３３，Ｓ１３４，Ｓ１３５の処理を行なう。フラットではない、すなわち、グラデーションの場合には、Ｓ１３６，Ｓ１３７，Ｓ１３８の処理を行なう。
【００６７】
対象のクラスタｃ_i の色分布がフラットである場合、まずＳ１３３で色分布がフラットであるクラスタのリストを作成する。ここでは、対象のクラスタｃ_i をフラットリストに追加する。この例では、効率化のために色分布がフラットのクラスタと特殊クラスタは１つのプロファイルを用いて色変換を行なうこととする。Ｓ１３３の処理はこのための前処理である。
【００６８】
Ｓ１３４では、対象のクラスタｃ_i が色再現範囲外の画素を含んでいるか否かを調べる。もし、含んでいれば、Ｓ１３５において、対象のクラスタｃ_i の代表色Ｍ_i に対してあらかじめ設定されている初期プロファイルを適用して代表色の色変換を行なう。この色変換の結果は、フラットリストに記録しておく。
【００６９】
この例における初期プロファイルは、色再現範囲外の色のみを明度および色相を変化させずに彩度方向に移動して色再現範囲の外縁に張り付ける変換ベクトルの集合により構成されているものとする。色再現範囲内の色については何ら変換を行なわない。図９は、色分布がフラットのクラスタｃ_i の代表色Ｍ_i を初期プロファイルに基づいて色変換する場合の説明図である。この図では、簡単にするために、彩度および色相で表わされる平面を示している。図中、実線は色再現範囲の外縁であり、黒丸はフラットのクラスタｃ_i の代表色Ｍ_i であり、実線の矢印は代表色Ｍ_i に対応する変換ベクトルｕ_i である。図９に示すように、例えば色再現範囲外の代表色Ｍ₂,Ｍ₅,Ｍ₆ は、変換ベクトルｕ₂,ｕ₅,ｕ₆ によって色再現範囲の外縁に変換される。このとき、色相、明度は保持され、彩度のみを変化させている。なお、色再現範囲内の代表色Ｍ₁,Ｍ₃,Ｍ₄,Ｍ₇,Ｍ₈ は変換しない。
【００７０】
対象のクラスタｃ_i の色分布がグラデーションである場合には、Ｓ１３６では、そのクラスタｃ_i が占有する画像平面上の領域をグラデーションマスクに追加する。最終的には、色分布がグラデーションであるすべてのクラスタが占有する領域に相当するマスクが作成される。これは、グラデーション部分の色変換をクラスタごとに行ない、その他の部分の色変換は後でまとめて行なうので、その他の部分の色変換の対象となるべき領域を得るために作成する。
【００７１】
Ｓ１３７では、対象のクラスタｃ_i が色再現範囲外の画素を含んでいるか否かを調べる。もし、含んでいなければ、特に色変換範囲内への変換は必要ないため、何もせずにＳ１３９の処理に進む。含んでいれば、Ｓ１３８において、グラデーションの色分布を持つクラスタｃ_i に対して階調性を保持した色再現範囲内への色変換を行なう。ここでは、対象のクラスタｃ_i に属する色再現範囲外の画素と色再現範囲の外縁との距離が最も大きなベクトルｖ_i を求めて、このベクトルｖ_i を変換ベクトルとして色再現範囲外の画素を含む色分布がグラデーションのクラスタに属する画素の色を変換する。
【００７２】
図１０は、色再現範囲外の色を含む色分布がグラデーションであるクラスタｃ_i に属する画素の色を色再現範囲内へ色変換する場合の説明図である。この図では、簡単にするために、彩度および色相で表わされる平面を示している。図中、実線は色再現範囲の外縁であり、黒丸はクラスタｃ_i に属する画素の色であり、実線の矢印はベクトルｖ_i である。また、点線の矢印はクラスタｃ_i に属するある画素の色をベクトルｖ_i で変換する様子を示しており、黒い四角は色変換後の画素の色を示している。色再現範囲の外縁との距離が最も大きなベクトルｖ_i を用いて色変換を行なうことによって、クラスタｃ_i 内の全ての色が色再現範囲内となる。このとき、クラスタｃ_i 内の各色間の相対的な変化量はそのまま保存されているので、階調性が保持される。
【００７３】
この例では、クラスタｃ_i に属するすべての画素の色をベクトルｖ_i により変換しているが、本発明ではこれに限定するものではなく、対象のクラスタに属する画素の色をクラスタ内の階調性を保持したまま色再現範囲内に変換できればどのような方法を用いてもよい。例えば、色変換による他の色への影響をできるだけ小さくするために、クラスタｃ_i に属するすべての画素の色の分散をベクトルｖ_i の大きさに応じて縮小した上で、色再現範囲内に色変換するように構成してもよい。
【００７４】
Ｓ１３９では、未処理のクラスタがまだ残っているか否かを調べて、残っていれば、Ｓ１３１に戻って処理を繰り返す。残っていなければ、Ｓ１４０、Ｓ１４１の処理を行なう。Ｓ１４０の処理を行なう直前には、フラットの色分布を持つクラスタのリストであるフラットリストが完成している。なお、Ｓ１３５の処理により色変換後の代表色もこのフラットリストに記録されており、色再現範囲外の代表色Ｍ_i はＭ_i'に変換されるものとする。色再現範囲内の代表色Ｍ_i については、その変換後の代表色Ｍ_i'と等しい。また、グラデーションの色分布を持つクラスタについてはＳ１３８の処理により既に色変換は完了している。
【００７５】
図１１は、フラットリストの一例の説明図である。この図は、図９に示した例を元に作成されている。したがって、代表色が色再現範囲内に存在するクラスタｃ₁,ｃ₃,ｃ₄,ｃ₇,ｃ₈ の代表色はその変換後の代表色と等しい。また、クラスタｃ₂,Ｃ₅,Ｃ₆ の代表色は色再現範囲外に存在するので、その色再現範囲外の代表色Ｍ₂,Ｍ₅,Ｍ₆ が変換された代表色Ｍ₂'，Ｍ₅'，Ｍ₆'は、色再現範囲の外縁の色である。
【００７６】
Ｓ１４０では、フラットリストを参照して、色変換後の代表色Ｍ_i'を走査することにより、色再現範囲外のクラスタの色変換後の代表色に類似する代表色を持つクラスタが存在するか否かを調べる。もし、これが存在すれば、色変換によって他の色との識別性が低下する可能性がある。このような色識別性の低下を防ぐために、初期プロファイルを修正する。ここでは、色再現範囲外のクラスタの色変換後の代表色Ｍ_i'の色相との差があらかじめ定められた閾値Ｔｈ₉ よりも小さいクラスタを類似する可能性があるとして抽出する。
【００７７】
図１２は、色再現範囲外のクラスタの色変換後の代表色に類似する代表色を持つクラスタが存在するか否かを調べる方法の一例の説明図である。この図では、色再現範囲外のクラスタの代表色Ｍ₂ について、その色相Ｈ₂ との色相の差がＴｈ₉ よりも小さい代表色を持つクラスタを探索する。この例では、代表色Ｍ₁ とＭ₃ が色相Ｈ₂ との色相の差がＴｈ₉ よりも小さい。そのため、代表色Ｍ₂ を持つクラスタを色変換すると、代表色Ｍ₁ とＭ₃ を持つクラスタと類似してしまう可能性があるものとして、この代表色Ｍ₁ ，Ｍ₃ が抽出される。
【００７８】
そして、類似する可能性があるとして抽出された代表色と、変換後の代表色との色差を算出して、この色差があらかじめ定められた閾値Ｔｈ₁₀よりも小さい場合に、色識別性が低下すると判定する。本発明では、この方法に限定するものではなく、色の識別性が低下する程度に類似した色の代表色を抽出することができる方法であれば、どのような方法を用いてもよい。
【００７９】
色識別性が低下すると判定された場合には、その原因となる色再現範囲外のクラスタの代表色もしくはこの近傍の色に対応する初期プロファイル中の変換ベクトルを、色識別性が低下しないように修正する。ここでは、色再現範囲外のクラスタの色変換後の代表色Ｍ_i'に類似している可能性があるとして抽出された１つ以上の代表色Ｍ_j との色差がＴｈ₁₀以上で、かつ、色変換後の代表色Ｍ_i'の色相を変化させずにＭ_i'との色差が最も小さな代表色Ｍ_i"を算出して、Ｍ_i からＭ_i"への変換ベクトルに置き換えることで修正を行なう。
【００８０】
図１３は、色識別性が低下すると判定された場合に代表色Ｍ_i"を算出する処理の一例の説明図である。図中、Ｍ₂ は色再現範囲外の色であり、Ｍ₂'は色再現範囲の外縁上の点、Ｍ₁ とＭ₃ はＭ₂'に類似する可能性があると判断された色、Ｍ₃ はさらに識別性が低下すると判定された色である。また、図１３は３次元の空間を表わしており、実線は色再現範囲の外縁と色相がＨ₂ である空間内の平面との交線を示している。さらに、破線は各色Ｍ₁ ，Ｍ₃ から半径Ｔｈ₁₀の球面を示している。ここで求めるＭ₂"は、色再現範囲の外縁と色Ｍ₃ を中心とする半径Ｔｈ₁₀の球面との交点である。こうして、Ｍ₂ からＭ₂"への変換ベクトルｕ₂'が算出でき、初期プロファイルの該当する変換ベクトルと置き換えられる。
【００８１】
このようにして、色識別性の低下を防ぐための修正を初期プロファイルに施し、グラデーションの色分布を持つクラスタ以外の部分の色変換を行なうプロファイルを作成する。
【００８２】
Ｓ１４１では、Ｓ１３６で生成されたグラデーションマスク以外の画素に対して、Ｓ１４０で作成したプロファイルを適用して、色変換を行なう。
【００８３】
このように、クラスタ別色変換部１７で行なう色変換処理は、クラスタ単位で行なわれる。上述の例のように重要色に限定したクラスタを対象とすることによって処理の効率化が実現でき、得られる結果も色変換により画像全体の再現性の変化を抑えたものとなる。
【００８４】
図２に戻り、出力側色変換部１８は、クラスタ別色変換部１７で色再現範囲内の色に変換された、内部処理で用いた表色系の画像データを、出力装置が出力可能な表色系の画像データに変換する。この変換は、既存の方法により行なうことができる。この例では、Ｌ^* Ｃ^* Ｈ゜表色系を内部処理で用いたため、Ｌ^* Ｃ^* Ｈ゜画像データを出力装置が出力可能な例えばＹＭＣＫ表色系の画像データに変換する。画像データ出力装置１９は、出力側色変換部１８が生成した、出力装置が出力可能な画像データを出力する。
【００８５】
このようにして、出力装置が再現できない色を含む画像データが入力された場合でも、グラデーション部分では色の階調性を保存し、またフラットな部分では識別性を保存して、再現可能な色に変換して出力することによって、良好な再現性を得ることができる。このとき、重要色について特にこのような変換を行なうことによって、処理の効率化とともに、画像全体の再現性を向上させることができる。
【００８６】
図１４は、本発明の画像処理装置および画像処理方法を実現する第２の実施の形態を示す概念ブロック図である。図中、図１と同様の部分には同じ符号を付して説明を省略する。６は重要領域作成部、７はグループ作成部、８は色変換部である。重要領域作成部６は、画像中のなるべく色を変化させたくない領域についての指示を受け、その指示に従って重要領域を作成する。グループ作成部７は、画像データ入力部１で入力された画像データから、重要領域作成部６で作成された重要領域と画像データ解析部２で抽出された重要色に基づいて、色およびその分布が類似するグループを作成する。色変換部８は、グループ作成部３で生成されたグループごとに、グループ内の色とその分布、重要領域作成部６で作成された重要領域、画像データ解析部２で抽出された重要色、および色再現範囲に基づいて色変換ベクトルを決定し、画像データ入力部１で入力された画像データの色変換を行なう。
【００８７】
画像データ入力部１で入力された出力対象の画像データは、量子化されて画像データ解析部２、グループ作成部３、色変換部４に与えられる。一方、画像中のなるべく色を変化させたくない領域に関する指示が入力され、重要領域作成部６において重要領域が作成される。画像データ解析部２は、与えられた画像データから重要と判断する１つ以上の色を重要色として抽出し、この重要色と、重要領域作成部６で作成した重要領域に基づいて、グループ作成部７は、与えられた画像データから色およびその分布が類似しているグループを生成する。色変換部８は、生成されたグループおよびグループ内の色分布、重要領域、重要色と、出力装置の色再現範囲に基づいて色変換ベクトルを作成し、これを与えられた画像データに適用して色変換を行なう。色変換後の画像データは、画像データ出力部５から出力装置に出力される。
【００８８】
図１５は、本発明の第２の実施の形態を実現する構成例を示すブロック図である。図中、図２と同様の部分には同じ符号を付して説明を省略する。２１は重要領域作成部、２２はクラスタ抽出部、２３はクラスタ別色変換部である。重要領域作成部２１は、入力画像の中で、色変換により画像全体としては大きく変化しないが、できるだけ色を変化しないことが好ましい領域の指示を受け取り、この領域を重要領域として一時データ記憶装置１３に記録する。ここでは、入力画像を表示し、マウスなどのポインティングデバイスを用いて閉領域を指定し、この閉領域を重要領域とする。特に必要がなければ、これらの重要領域は指定しなくてもよい。本発明は、ここで説明した方法に限定するものではなく、できるだけ色を変化しないことが好ましい領域を得られる方法であれば、どのような方法を用いてもよい。例えば、画像を入力する際に、これらの領域を表わすような二値画像を読み込むように構成することもできる。この例では、作成された重要領域は正確に目的の領域のみを含んでいると仮定するが、目的の領域を大まかに指定してその領域の境界近辺を調査することで正確な領域を得るように構成することもできる。
【００８９】
クラスタ抽出部２２は、重要領域作成部２１で作成した重要領域と重要色抽出部１４で抽出した重要色に基づいて類似する特徴を持つ画素の集合であるクラスタを抽出する。例えば重要領域に含まれる画素と重要色もしくは重要色に近い色を持つ画素のみからクラスタを構成することにより、効率的にクラスタの抽出が実現できる。この例では、重要色と大きく異なる重要領域以外の色については、色変換によって画像全体の色再現性は大きく変化しないとみなして色識別性や階調性の保存のための特別な処理は行なわない。
【００９０】
この例でも、Ｌ^* Ｃ^* Ｈ゜色空間における明度、彩度、色相と画像平面におけるＸ座標、Ｙ座標の５次元の特徴空間上での再配置法によるクラスタリングによりクラスタを抽出する。クラスタ抽出部２２の動作の一例としては、上述の第１の実施の形態と同様、図４に示したフローチャートに従って処理を行なえばよい。この第２の実施の形態では、図４のＳ１０１で初期クラスタを生成する際に、重要色との色差が閾値Ｔｈ₂ よりも小さい画素だけでなく、重要領域自体も初期クラスタとする。こうすることにより、重要領域に含まれる画素も、抽出されるクラスタに属するため、クラスタ解析部１６およびクラスタ別色変換部２３を経て、重要領域の階調性および色識別性を保持した色変換が実現できる。
【００９１】
この例では、重要色と重要領域の重要性を同等と見なしてこのような処理を行なうが、重要色と重要領域の重要性に差を付けて、例えば、重要領域の重要性を重視すれば重要領域以外の画像中の画素を対象にクラスタ抽出部１５の処理を行ない、これに１つの重要領域を１つのクラスタとして追加するように構成することもできる。また、不必要に重要領域が分割されたり他の領域と統合されたり、さらに削除されたりしないように、クラスタに対して重要領域であるという属性を付加しておくとよい。さらに、重要領域にレベルを設けてそのレベルによって処理を分けて行なってもよい。
【００９２】
クラスタ別色変換部２３は、上述の第１の実施の形態におけるクラスタ別色変換部１７と同様に構成することも可能であるが、ここでは、クラスタごとにその重要度Ｉ_i を定義して、重要度の高いクラスタの画素の色はなるべく変換しないように構成する例について説明する。
【００９３】
クラスタの重要度は、クラスタｃ_i に属する画素の総数Ｎｃ_i や重要領域の画素数、重要色と代表色との色差などを用いて定義することができる。この例では、クラスタｃ_i に属する画素の総数Ｎｃ_i をそのままクラスタの重要度として用いることにする。したがって、クラスタｃ_i に属する画素の総数Ｎｃ_i が多いほど重要なクラスタとみなす。
【００９４】
クラスタ別色変換部２３における処理は、上述の第１の実施の形態におけるクラスタ別色変換部１７と同様であり、図８に示すフローチャートに従って処理を行なえばよい。ただし、この第２の実施の形態では重要度を用い、色再現範囲外に存在する代表色を持つ色分布がフラットのクラスタを色再現範囲内に色変換する際に、色識別性が低下すると判定された後の変換ベクトルの修正方法を変更する。
【００９５】
上述の第１の実施の形態では、図１３において説明したように、色識別性が低下すると判定された場合、一律に、色再現範囲外に存在する代表色を持つ、色分布がフラットのクラスタを変換する変換ベクトルを修正していた。この第２の実施の形態では、クラスタの重要度を用い、重要度の低いクラスタの変換ベクトルを修正するように構成する。したがって、色再現範囲外に存在する代表色を持つ、色分布がフラットのクラスタの方が、色識別性が低下する色再現範囲内のクラスタよりも重要度が低い場合には、図１３に示したように色再現範囲外に存在する代表色を持つ、色分布がフラットのクラスタを変換する変換ベクトルを修正するが、重要度が高い場合には、色識別性が低下する色再現範囲内のクラスタの代表色を、色識別性が低下しない色にまで変換する。
【００９６】
図１６は、色再現範囲内のクラスタの代表色を色識別性が低下しないように変換する変換ベクトルを算出する処理の一例の説明図である。図中、Ｍ₂ は色再現範囲外の色であり、Ｍ₂'は色再現範囲の外縁上の点、Ｍ₁ とＭ₃ はＭ₂'に類似する可能性があると判断された色、Ｍ₃ はさらに色識別性が低下すると判定された色である。また、図１６は３次元の空間を表わしており、実線は色再現範囲の外縁と色相がＨ₂ である空間内の平面との交線を示している。さらに、破線は半径Ｔｈ₁₀の球面を示している。
【００９７】
ここで、代表色Ｍ₂ を持つクラスタの重要度と、代表色Ｍ₃ を持つクラスタの重要度を比較すると、代表色Ｍ₂ を持つクラスタの方が重要度が高いとする。この場合、図１３に示したように代表色Ｍ₂ の変換ベクトルｕ₂ を変更させるのではなく、図１６に示すように、色識別性が低下すると判定された色Ｍ₃ を移動させる。例えば、色Ｍ₃ を中心とする半径Ｔｈ₁₀の球を移動して、Ｍ₂'を接点に持ち、かつ、Ｍ₃ との色差が最も小さな色Ｍ₃'を算出する。そして、Ｍ₃ をＭ₃'に変換する変換ベクトルｕ₃ を求めて、この変換ベクトルにより初期プロファイルの対応する変換ベクトルと置き換える。このように構成することにより、画像中で色変換によってより大きな再現性の変化を生じる色ほど、変化の幅を小さくすることができる。
【００９８】
【発明の効果】
以上の説明から明らかなように、本発明によれば、入力された画像データに応じて色分布をグループ化して、グループごとに色再現範囲内となるように変換ベクトルを決定するので、画像全体の色を大きく変化させることなく、色識別性と階調性の低下を防ぐことができる。さらに、重要色、あるいは重要色と重要色に類似する色に限定したグループ化を行なうことによって、効率的に色再現範囲内への色変換を行なうことができる。さらに、色を変化させることで画像全体の再現性の変化する度合いを重要度として導入することで、より入力画像に忠実な色の再現が行なえるように、色再現範囲内への色変換が実現できるという効果がある。また、画像中における比較的小さな部分でも、その領域を重要領域として指定することでグラデーションなどの階調性を保持することができる。
【図面の簡単な説明】
【図１】 本発明の画像処理装置および画像処理方法を実現する第１の実施の形態を示す概念ブロック図である。
【図２】 本発明の第１の実施の形態を実現する構成例を示すブロック図である。
【図３】 Ｌ^* Ｃ^* Ｈ°色空間を分割して作成する単位空間の説明図である。
【図４】 本発明の第１の実施の形態を実現する構成例におけるクラスタ抽出部１５の動作の一例を示すフローチャートである。
【図５】 本発明の第１の実施の形態を実現する構成例におけるクラスタ解析部１６のクラスタの色分布認識処理の動作の一例を示すフローチャートである。
【図６】 本発明の第１の実施の形態を実現する構成例におけるクラスタ解析部１６のクラスタの統合と削除処理の動作の一例を示すフローチャートである。
【図７】 本発明の第１の実施の形態を実現する構成例においてクラスタ別色変換部１７が行なう色変換とクラスタの色分布との対応の一例の説明図である。
【図８】 本発明の第１の実施の形態を実現する構成例におけるクラスタ別色変換部１７の動作の一例を示すフローチャートである。
【図９】 色分布がフラットのクラスタｃ_i の代表色Ｍ_i を初期プロファイルに基づいて色変換する場合の説明図である。
【図１０】 色再現範囲外の色を含む色分布がグラデーションであるクラスタｃ_i に属する画素の色を色再現範囲内へ色変換する場合の説明図である。
【図１１】 フラットリストの一例の説明図である。
【図１２】 色再現範囲外のクラスタの色変換後の代表色に類似する代表色を持つクラスタが存在するか否かを調べる方法の一例の説明図である。
【図１３】 色識別性が低下すると判定された場合に代表色Ｍ_i"を算出する処理の一例の説明図である。
【図１４】 本発明の画像処理装置および画像処理方法を実現する第２の実施の形態を示す概念ブロック図である。
【図１５】 本発明の第２の実施の形態を実現する構成例を示すブロック図である。
【図１６】 色再現範囲内のクラスタの代表色を色識別性が低下しないように変換する変換ベクトルを算出する処理の一例の説明図である。
【符号の説明】
１…画像データ入力部、２…画像データ解析部、３…グループ作成部、４…色変換部、５…画像データ出力部、６…重要領域作成部、７…グループ作成部、８…色変換部、１１…画像データ入力装置、１２…入力側色変換部、１３…一時データ記憶装置、１４…重要色抽出部、１５…クラスタ抽出部、１６…クラスタ解析部、１７…クラスタ別色変換部、１８…出力側色変換部、１９…画像データ出力装置、２１…重要領域作成部、２２…クラスタ抽出部、２３…クラスタ別色変換部。[0001]
BACKGROUND OF THE INVENTION
The present invention relates to an image processing apparatus and an image processing method for performing color conversion on input color image data so as to be within a color reproduction range of an output device (device).
[0002]
[Prior art]
As one of the image processes, there is a color conversion process for performing color conversion of color image data so as to be within the color reproduction range of an apparatus (device) that outputs color image data. For example, when a color image is output on a display, a printer, or the like, not all colors can be expressed, and only colors within the color reproduction range of the output device (device) are output. Hereinafter, the assumed color reproduction range of the output device (device) is simply referred to as a color reproduction range. When the color image data to be output has a color that deviates from the color reproduction range, color conversion processing for converting a color that deviates from the color reproduction range into a reproducible color is required.
[0003]
Conventionally, various methods have been devised for color conversion of how to reproduce colors that cannot be reproduced by an output device. There are three typical conversion methods: a conversion method that preserves gradation, a conversion method that preserves hue, and a conversion method that preserves brightness.
[0004]
The conversion method that preserves gradation is a method that changes not only colors outside the color reproduction range but also reproducible colors at a certain ratio so that all colors of the target color image data can be reproduced. . In this method, the difference in each color is preserved, so that the color change such as gradation is preserved. Therefore, it is suitable for color conversion for an image such as a photograph that emphasizes gradation. However, when the target color image has a color greatly deviating from the color reproduction range, there is a drawback that the color of the entire color image changes greatly.
[0005]
On the other hand, in the conversion method that preserves the hue and lightness, only the color outside the color reproduction range is changed to the outer edge color of the color reproduction range without changing the hue and lightness. As a result, the overall color change can be kept low. However, since only the color outside the color reproduction range is changed, if the converted color and its neighboring colors exist in the color image data before the conversion, the distinguishability between those colors and the changed colors is increased. There is a disadvantage that it deteriorates. For example, if a part of the gradation is a color outside the color reproduction range, this conversion causes a part of the gradation to be the same color, resulting in a reduction in gradation. In addition, the color that could be identified in the color image data before the conversion becomes almost the same color due to this conversion, and the distinguishability may deteriorate.
[0006]
In order to compensate for the drawbacks of the three typical conversion methods described above, there is also a method of selecting a conversion method according to the type and characteristics of the image. For example, in the device described in Japanese Patent Laid-Open No. 5-16739, the number of pixels outside the color reproduction range is counted, and in the device described in Japanese Patent Laid-Open No. 4-287469, the number of pixels in the color reproduction range. When the ratio of the number of pixels outside the color reproduction range is small, color conversion is performed using a conversion method that preserves the hue, and when the ratio of the number of pixels within the color reproduction range is small, the gradation is preserved. Color conversion is performed by the conversion method. Furthermore, in the apparatus described in Japanese Patent Laid-Open No. 6-162181, the color space is divided into a plurality of blocks for each hue, the number of pixels outside the color reproduction range is counted for each block, and the ratio of the count values is When there is a large block, color conversion is performed using a conversion method that preserves gradation, and in other cases, color conversion is performed using a conversion method that preserves hue. In these apparatuses, since the conversion method is selected based on the number of pixels inside and outside the color reproduction range, there is a problem that a part of gradation of gradation including many colors within the color reproduction range is lost.
[0007]
Further, for example, in the color image conversion apparatus described in Japanese Patent Application Laid-Open No. 7-203234, the uniform color space is divided into unit spaces, the number of pixels included in each unit space is obtained, and is out of the color reproduction range. For the unit area, color conversion is performed by mapping the unit area to another color unit area having a constant brightness according to the number of pixels. According to this technology, since the case where a color outside the color reproduction range is converted to the same color as other used colors is reduced, the gradation of gradation can be preserved, and the color distinctiveness can be preserved. The decrease can also be kept low. In addition, since colors within the color reproduction range are almost always expressed as they are, unnecessary color changes can be prevented. However, since color conversion is performed only for a unit space outside the color reproduction range, for example, color gradation may be lost due to color conversion in the unit space outside the color reproduction range.
[0008]
In order to solve the problems described so far, for example, in Japanese Patent Application No. 8-118026, attention is paid to the partial color distribution and position of the color image, and the gradation of the gradation portion and the color identification of the flat portion. An image processing apparatus that retains the characteristics and does not degrade the color reproducibility of the entire color image as much as possible is shown. This method creates groups for each color distribution and position in the color image and performs color conversion for each group. According to this method, the gradation part can be reproduced as a gradation, and at the same time the flat part is The entire color image can be reproduced as faithfully as possible while maintaining color discrimination. However, since a group is created for each color distribution and position of color image data, there is a drawback that the processing speed is slow. In addition, since color conversion is performed so that the gradation property of the gradation portion and the color discrimination property of the flat portion are equally maintained for all colors included in the color image data, it has a great influence on the color image. The color may be changed unnecessarily.
[0009]
[Problems to be solved by the invention]
The present invention has been made in view of the above-described circumstances, and is determined to be important in an image while maintaining the gradation of a gradation portion and the color discrimination of a flat portion existing in the same image. It is an object of the present invention to provide an image processing apparatus and an image processing method for efficiently performing color conversion processing with particular attention to colors and regions.
[0010]
[Means for Solving the Problems]
  According to the first aspect of the present invention, in the image processing apparatus, image data input means for inputting image data, and image data analysis for extracting one or more important colors from the image data input by the image data input means Recognizing the color of the image data and its distribution on the image plane based on the important color extracted by the image data analyzing means and the color similar to the important color and different colors on the image plane Group creating means for generating a group for each distribution, group analyzing means for recognizing the color distribution for each group generated by the group creating means and classifying the group into gradation or flat, and the image data analyzing means Color conversion according to the classification by the group analysis means based on the important color extracted by the color and the color reproduction range of the output device. It is characterized in that it comprises a power sale color conversion means.
[0011]
  According to a second aspect of the present invention, in the image processing apparatus, image data input means for inputting image data, area instruction means for instructing an important area in the image data input by the image data input means, and Image data analysis means for extracting one or more important colors from the image data input by the image data input means, the important colors extracted by the image data analysis means, colors similar to the important colors, and the region indication Group creating means for recognizing the color of the image data and its distribution on the image plane based on the important area instructed by the means and generating a group for each different color and distribution on the image plane; Group analysis means for recognizing the color distribution of each group generated by the creation means and classifying the group into gradation or flat , It is characterized in that it comprises a color conversion unit that performs color conversion according to the classification by the group analysis unit based on the color reproduction range of the image data analysis means important color output device extracted by.
[0012]
According to a third aspect of the present invention, in the image processing apparatus according to the first or second aspect, the image data analyzing unit calculates a frequency in the color space of the image data input by the image data input unit. A color having a frequency or a ratio of the frequency to the whole of which is a threshold value or more is extracted as an important color.
[0013]
According to a fourth aspect of the present invention, in the image processing apparatus according to the third aspect, the image data analyzing unit is configured such that the image data input by the image data input unit is a color outside a color reproduction range in the output device. The important color is passed to the group creating means.
[0014]
According to a fifth aspect of the present invention, in the image processing apparatus according to the first or second aspect, the group creating unit creates a group by clustering based on the important color extracted by the image data analyzing unit. It is what.
[0015]
According to a sixth aspect of the present invention, in the image processing apparatus according to the fifth aspect, the group creating means includes a clustering threshold value between a cluster including the important color extracted by the image data analyzing means and a cluster not including the important color. It is characterized by changing parameters.
[0016]
According to a seventh aspect of the present invention, in the image processing apparatus according to the first or second aspect, the color conversion unit determines a color conversion parameter so as not to change the important color extracted by the image data analysis unit as much as possible. The color conversion is performed.
[0017]
According to an eighth aspect of the present invention, in the image processing apparatus according to the second aspect, the color conversion means includes an important color extracted by the image data analysis means and an important area designated by the area instruction means. In this case, the color conversion parameters are determined so as not to change the color and gradation in the region as much as possible, and color conversion is performed.
[0018]
  According to the ninth aspect of the present invention, in the image processing method, one or more important colors are extracted from the input image data by the image data analysis means, and the extracted important colors and colors similar to the important colors are extracted. Based on the color of the image data and its distribution on the image plane, a group for each different color and its distribution on the image plane is generated by the group creating means, and the generated color distribution for each group is generated. Recognizing and classifying the group into gradation or flat by group analysis means, and performing color conversion according to the classification by the color conversion means based on the important color and the color reproduction range of the output device. is there.
[0019]
  According to a tenth aspect of the present invention, in the image processing method, one or more important colors are received from the input image data while receiving an instruction about an important area in the image data input by the image data input means. Are extracted by image data analysis means, and the color of the image data and its distribution on the image plane are recognized on the basis of the extracted important color, the color similar to the important color and the important area, and different colors A group for each distribution on the image plane is generated by the group creating means, the generated color distribution for each group is recognized, the group is classified into gradation or flat by the group analyzing means, and the important colors and The color conversion unit performs color conversion according to the classification based on the color reproduction range of the output device.
[0020]
According to an eleventh aspect of the present invention, in the image processing method according to the ninth or tenth aspect, when the important color is extracted, a frequency in the color space of the image data is calculated, and the frequency or the frequency is calculated. A color whose ratio to the whole is equal to or greater than a threshold is extracted as an important color.
[0021]
According to a twelfth aspect of the present invention, in the image processing method according to the ninth or tenth aspect, when the important color is extracted, a color outside the color reproduction range in the output device of the input image data and A color that is estimated to convert a color outside the color reproduction range is extracted as an important color.
[0022]
According to a thirteenth aspect of the present invention, in the image processing method according to the ninth or tenth aspect, when the group is generated, the group is generated by clustering based on the important color. .
[0023]
The invention according to claim 14 is the image processing method according to claim 13, wherein the clustering is performed by changing a clustering threshold value or a parameter between a cluster including the important color and a cluster not including the important color. To do.
[0024]
According to a fifteenth aspect of the present invention, in the image processing method according to the ninth or tenth aspect, when performing the color conversion, a color conversion parameter is determined so as not to change the important color as much as possible, and the color conversion is performed. It is characterized by doing.
[0025]
According to a sixteenth aspect of the present invention, in the image processing method according to the tenth aspect, when the color conversion is performed, if the important color and the designated important area exist, the color and the gradation in the area are indicated. Color conversion parameters are determined so that the tonality is not changed as much as possible, and color conversion is performed.
[0026]
DETAILED DESCRIPTION OF THE INVENTION
FIG. 1 is a conceptual block diagram showing a first embodiment for realizing an image processing apparatus and an image processing method of the present invention. In the figure, 1 is an image data input unit, 2 is an image data analysis unit, 3 is a group creation unit, 4 is a color conversion unit, and 5 is an image data output unit.
[0027]
The image data input unit 1 quantizes the input image and generates image data. The image data analysis unit 2 extracts one or more important colors determined to be important colors from the image data generated by the image data input unit 1. The group creation unit 3 creates groups having similar colors and distributions based on the important colors extracted by the image data analysis unit 2 from the image data input by the image data input unit 1. The color conversion unit 4 determines a color conversion vector for each group generated by the group creation unit 3 based on the colors in the group and their distribution, the important colors extracted by the image data analysis unit 2, and the color reproduction range. Then, color conversion of the image data input by the image data input unit 1 is performed. The image data output unit 5 outputs the image data color-converted by the color conversion unit 4 to an output device (device).
[0028]
The output target image data input by the image data input unit 1 is quantized and supplied to the image data analysis unit 2, group creation unit 3, and color conversion unit 4. In the image data analysis unit 2, one or more colors that are determined to be important are extracted from the given image data and provided to the group creation unit 3 and the color conversion unit 4. The group creation unit 3 generates a group having similar colors and distributions based on the important color from the given image data, and gives the group to the color conversion unit 4. The color conversion unit 4 performs color conversion by creating a color conversion vector based on the given group and the color distribution within the group, the important color, and the color reproduction range and applying it to the given image data. The image data after color conversion is output from the image data output unit 5 to an output device (device).
[0029]
FIG. 2 is a block diagram showing a configuration example for realizing the first embodiment of the present invention. In the figure, 11 is an image data input device, 12 is an input side color conversion unit, 13 is a temporary data storage device, 14 is an important color extraction unit, 15 is a cluster extraction unit, 16 is a cluster analysis unit, and 17 is a color conversion for each cluster. Reference numeral 18 denotes an output side color conversion unit, and 19 denotes an image data output device.
[0030]
The image data input device 11 reads an image and outputs it as input image data. For example, it can be configured by a digital camera, a scanner, or the like. Here, it is assumed that the image data is captured as image data in the RGB color space. Note that image data read in advance and stored in a storage device such as a hard disk, or image data transmitted via a network or the like may be used as input image data. Also, image data drawn using a computer may be used as input image data.
[0031]
The input side color conversion unit 12 performs color conversion between the color system of the input image data and the color system used in the internal processing. In this example, L in internal processing^*C^*Use the H color system. Therefore, the input side color conversion unit 12 uses the known method to calculate the L color from the RGB color space.^*C^*Conversion to the H ° color space is performed. Thereafter, L^*C^*Input image data expressed by the H color system is simply referred to as image data. Here L^*C^*Although the H ° color system is used, the present invention is not limited to this, and other color systems such as CIE L^*a^*b^*A color system may be used. In this case, the input side color converter 12 converts the RGB color space to L^*a^*b^*Performs conversion to color space.
[0032]
The temporary data storage device 13 is composed of a storage device such as a hard disk or a memory, for example. Store some data.
[0033]
The important color extraction unit 14 extracts a color having a high frequency in the input image as a color that is considered to have a large change in the reproducibility of the input image due to the color conversion, and sets this as an important color. Here, first, the division information set in advance, for example, the division number Np for each axis of the color space is used to calculate the L^*C^*A unit space having a lattice point formed by dividing the H ° color space at the center of gravity is generated. Then, by counting the number of pixels for each unit space of the image data, a three-dimensional histogram based on brightness, saturation, and hue is created. And a preset threshold value Th₁It is possible to extract the center of gravity of the unit space having a greater frequency as an important color. Here, one important color is a three-dimensional vector A composed of three elements of brightness, saturation, and hue._i(1 ≦ i ≦ N_impN_impCan be expressed by the number of important colors). Here, a unit space is created and important colors are extracted according to the frequency of each unit space. However, the present invention is not limited to this, and any method can be used as long as it can extract colors with a high frequency in the image. Such a method may be used.
[0034]
3 shows L^*C^*It is explanatory drawing of the unit space created by dividing the H ° color space. FIG. 3A is an explanatory diagram of lattice points formed when the color space is divided, and FIG. 3B is an explanatory diagram of a unit space having the above-described lattice points at the center of gravity. FIG. 3A shows an example in which Np = 4, and the solid line in the figure shows a part of a plane obtained by dividing each axis. The intersections of the three axes and the dividing plane are lattice points, and in FIG. 3A, only a part of the outside is indicated by black circles.
[0035]
The black circles shown in FIG. 3B are grid points formed by dividing the color space, and a unit space having this grid point at the center of gravity is generated so as to be adjacent to the same shape and size as the adjacent unit space. To do. In FIG. 3B, four unit spaces having four adjacent lattice points at the center of gravity are indicated by solid lines and broken lines.
[0036]
Here, if the image data does not include a color outside the color reproduction range, the following processing of the cluster extraction unit 15, the cluster analysis unit 16, and the cluster-specific color conversion unit 17 is unnecessary. Therefore, using the three-dimensional histogram created by the important color extraction unit 14 and the color reproduction range stored in advance, it is checked whether or not the image data includes a color outside the color reproduction range. If no color is included, the cluster extraction unit 15, the cluster analysis unit 16, and the cluster-specific color conversion unit 17 may be skipped.
[0037]
The cluster extraction unit 15 extracts a cluster that is a set of pixels having similar characteristics based on the important color extracted by the important color extraction unit 14. Thus, efficient cluster extraction can be realized by forming a cluster only from pixels having important colors or colors close to important colors. In this example, regarding the color that is significantly different from the important color, it is considered that the reproducibility of the entire image is not greatly changed by the color conversion, and no special processing for storing the color discrimination and gradation is performed.
[0038]
In this example, as a feature used for the process of extracting a cluster, L^*C^*Five features were used: brightness, saturation, hue in the H ° color space, and X and Y coordinates in the image plane. The present invention does not limit the features to be used to these five, and any number of features may be used as long as the features to be used include features relating to color. For example, CIE L^*a^*b^*Three features represented in color space may be used.
[0039]
The process of extracting a cluster which is a set of pixels having similar features can be realized by a process generally called clustering. For example, the rearrangement method which is one of the non-hierarchical clustering shown in “Image Analysis Handbook”, University of Tokyo Press (hereinafter referred to as Document 1) can be used. The present invention is not limited to this rearrangement method, and any method may be used as long as it can extract clusters having similar characteristics.
[0040]
FIG. 4 is a flowchart showing an example of the operation of the cluster extraction unit 15 in the configuration example for realizing the first embodiment of the present invention. In this example, a pixel similar to the important color is set as the initial cluster.
[0041]
First, in S101, an initial cluster is generated based on the important color extracted by the important color extraction unit. Here, important color A_i(1 ≦ i ≦ N_impN_impIs the number of important colors) and the coordinates D of the image data [x, y] (1 ≦ x ≦ W; 1 ≦ y ≦ H; W and H are the width and height of the image data, respectively)._xy= (L_xy, C_xy, H_xy,) (L_xy, C_xy, H_xyAre the absolute values of the differences from the pixel brightness (brightness, saturation, hue) at the coordinates [x, y] of the image data, respectively.₂Smaller pixels are extracted as one initial cluster. As a result, one or more initial clusters consisting of one pixel can be extracted.
[0042]
In addition, a cluster center used for pixel rearrangement is calculated for each extracted initial cluster. Here, a five-dimensional vector V consisting of the brightness, saturation, hue and X-coordinate and Y-coordinate in the image plane of only one pixel belonging to each initial cluster._i (1 ≦ i ≦ Nc; Nc is the number of clusters) was defined as the cluster center. In the subsequent pixel rearrangement processing, only the pixels belonging to the cluster extracted here are the processing targets. Therefore, the pixel set to be processed is temporarily stored.
[0043]
In S102, in order to speed up the convergence of the loop after S103, the initial cluster composed of one pixel generated in S101 is integrated. In this example, the cluster center V of the two clusters generated in S101._i (1 ≦ i ≦ Nc) and V_jDistance d with (1 ≦ j ≦ Nc)_ijIs a preset threshold Th_Three If smaller than these, these clusters are integrated and the cluster center is recalculated. Cluster center V_i (1 ≦ i ≦ Nc) and V_j Distance d with (1 ≦ j ≦ Nc)_ijWas calculated using equation (1).
[Expression 1]

Where V_i = (L_i, C_i, H_i, x_i, y_i ), V_j = (L_j, C_j, H_j, x_j, y_j ).
[0044]
In addition, an average vector of five-dimensional vectors composed of the brightness, saturation, hue, and the X coordinate and Y coordinate in the image plan view of the pixel belonging to the cluster was used at the cluster center of the cluster composed of a plurality of pixels. For example, cluster center V_i (1 ≦ i ≦ Nc) can be calculated by using equation (2).
[Expression 2]

Where Nc_i Is cluster c_i Is the total number of pixels belonging to L_ik, C_ik, H_ik, X_ik, Y_ikRespectively, cluster c_i Are the lightness, saturation, hue, X coordinate, and Y coordinate of the kth pixel.
[0045]
In S103, for the pixels belonging to the cluster extracted in S101, it is checked which cluster center is closest to the cluster, and the cluster is reconfigured. Here, the weighted distances between the pixel to be rearranged and all cluster centers are calculated in a feature space consisting of five dimensions of pixel brightness, saturation, hue, and X and Y coordinates of the image plane. Here, in order to improve the efficiency of processing, the weighted distance may be calculated only for the cluster center existing in the vicinity of the pixel to be rearranged instead of all the clusters. Pixels are rearranged so that the target pixel belongs to the cluster corresponding to the cluster center having the smallest weighted distance. Weighted distance d_i For example, the target pixel and the cluster center V_i From (1 ≦ i ≦ Nc), it can be obtained by equation (3).
[Equation 3]

Where (L_T, C_T, H_T, x_T, y_T ) Are the coordinates on the 5-dimensional feature space of the target pixel T, and V_i = (L_i, C_i, H_i, x_i, y_i ). Further, a, b, c, d, and e are weighting factors corresponding to the respective features. By appropriately determining this weighting factor, the generated group can be adjusted. For example, if c is set relatively large and d and e are set relatively small, if the hue is close, it can be configured as one cluster even if the distance on the image plane is slightly longer.
[0046]
As a result of the above-described calculation of the weighted distance, the smallest weighted distance between a certain pixel and the cluster center is set to a predetermined threshold value Th._Four In the case of larger than this, in order to create a more appropriate cluster, it is also possible to newly generate a cluster of one pixel.
[0047]
In S104, the cluster center of the cluster reconfigured in S103 is calculated. The cluster center calculation method is the same as the method described in S102.
[0048]
In S105, it is checked whether or not the cluster has been changed by the pixel rearrangement performed in S103. If not changed, the pixel rearrangement is considered to have converged, and the cluster extraction process is terminated. If it has been changed, the process returns to S103 and the process is repeated.
[0049]
In the example described here, clusters were extracted in a form limited to important colors and colors similar to important colors, but the colors were classified into multiple classes according to the distance from the important colors, and rearranged according to the representative color class. It is also possible to change the parameters and thresholds used. For example, by defining three classes of important colors, colors similar to important colors, and other colors, and setting thresholds so that smaller clusters are obtained for classes closer to important colors, it is important to Color distribution as detailed as the color is obtained, and as a result, color conversion with higher reproducibility can be realized.
[0050]
Returning to FIG. 2, the cluster analysis unit 16 recognizes the color distribution in the cluster extracted by the cluster extraction unit 15, and performs division, integration, and deletion of the cluster based on this color distribution as necessary. In this example, the colors are classified into two color distributions, gradation and flat, depending on whether the colors of the pixels belonging to the cluster are dispersed or concentrated. Then, after integrating clusters having similar color distributions, the minute clusters are deleted.
[0051]
FIG. 5 is a flowchart showing an example of the operation of the cluster color distribution recognition process of the cluster analysis unit 16 in the configuration example for realizing the first embodiment of the present invention. In S111, the cluster c whose color distribution is not recognized_i Select one. Subsequent processing is performed for this cluster c._i To the target.
[0052]
In S112, the L of pixels belonging to the cluster is determined in order to determine the color distribution in the cluster.^*C^*A dispersion vector in the H ° color space is calculated. For example, by using the equation (4), the cluster c_i Variance vector S_i ²(1 ≦ i ≦ Nc) can be calculated.
[Expression 4]

Where cluster c_i Cluster center V_i = (L_i, C_i, H_i, x_i, y_i ). Thereafter, M_i = (L_i, C_i, H_i ) For cluster c_i This is called the representative color.
[0053]
In S113, L of pixels belonging to the cluster^*C^*Threshold value Th in which the magnitude of the dispersion vector in the H ° color space is set in advance_Five If it is smaller, it is determined that the color distribution in the cluster is substantially uniform, and the cluster c is determined in S114._i An attribute of “flat” is added to. If it is larger, it is determined that the color distribution in the cluster changes with a certain width, and the cluster c is determined in S115._i An attribute of “gradation” is added to. Of course, each variance value of brightness, saturation, and hue and the threshold value Th_FiveMay be compared, or a different threshold Th for each value of brightness, saturation, and hue._5L, Th_5C, Th_5H May be compared.
[0054]
In S116, it is checked whether or not there is a cluster whose color distribution has not yet been recognized. If it remains, the process returns to S111 and the process is repeated. If not, the color distribution recognition process ends.
[0055]
Here, only the dispersion vector is used to examine the color distribution of the cluster. However, the present invention is not limited to this, and the color changes smoothly from “flat” where the color distribution is substantially uniform. Any method can be used as long as it can be classified into two color distributions of “gradation”. Further, for example, following the above-described processing, the threshold Th in which the size of the dispersion vector is set in advance is set._Five Is larger than the threshold value Th that the color change rate is set in advance by further examining the color change rate of adjacent pixels in the cluster._FiveIt is also possible to add an attribute of “gradation” when it is smaller than “other” and “other” when it is larger.
[0056]
FIG. 6 is a flowchart showing an example of cluster integration and deletion processing operations of the cluster analysis unit 16 in the configuration example for realizing the first exemplary embodiment of the present invention. First, in S121, cluster c in which integration and deletion processes are not performed._i Select one. In S122, cluster c_i Cluster c that has not been compared with_jSelect one.
[0057]
In S123, cluster c_i And cluster c_j Representative color M_i And M_j Threshold value Th in which the magnitude of the difference vector is preset₆ If it is smaller, in step S124, the cluster c_i And cluster c_j Variance vector S_i ²And S_j ²Threshold value Th in which the magnitude of the difference vector is preset₇ Check if it is smaller than. If this is also small, two clusters c_i And cluster c_j Are determined to be similar to each other, and in S125, the cluster c_i Cluster c_j To integrate. When integrated, cluster c_j Is deleted and updated by integration c_i Is the cluster representative color M again_i And the variance vector S_i ²Is calculated in advance. If the condition is not satisfied during the comparison in S123 and S124, the process proceeds to S126.
[0058]
In S126, cluster c_i Cluster c to be compared with_j Is still left, and if it remains, the process returns to S122 and the process is repeated. If not, the process proceeds to S127.
[0059]
In S127, the cluster c to be processed_i Total number of pixels belonging to Nc_i Is preset and the threshold value Th₈ Check if it is smaller than. And if it is small, cluster c_i Is determined to be an isolated minute cluster, and is deleted in S128. If larger, the process proceeds to S129 as it is.
[0060]
In S129, the cluster c not selected in S121_i Is checked, if it remains, the process returns to S121 to repeat the process. If not, the cluster integration and deletion process is terminated.
[0061]
In this example, the cluster integration condition is that the representative color of the two clusters and the color distribution are similar. However, in the case of a gradation having a large range, there is a disadvantage that the cluster is divided into a plurality of clusters. . Therefore, when configuring a higher-accuracy color conversion device, if the color change rate of the adjacent portion of the cluster having the gradation color distribution is examined, and if this is small, the two clusters are integrated. Good.
[0062]
The color conversion unit 17 for each cluster in FIG. 2 creates a color conversion profile for each cluster based on the color distribution of the clusters recognized by the cluster analysis unit 16. Then, color conversion is performed by applying a corresponding profile to the pixel set belonging to each cluster. At this time, it is not necessary to create a different profile for every cluster, and if there is no inconvenience even if the same profile is used, color conversion can be performed using one profile for a plurality of clusters. . Here, the profile is a set of conversion vectors for converting a certain color into a certain color, and color conversion can be performed based on the conversion vector described in the profile by applying the profile to the image. it can. At this time, since it is generally difficult to define conversion vectors for all colors, in general, when there is no conversion vector corresponding to the color to be converted, a plurality of neighboring conversion vectors are used to change the color. Perform conversion.
[0063]
FIG. 7 is an explanatory diagram showing an example of the correspondence between the color conversion performed by the cluster-specific color conversion unit 17 and the color distribution of the cluster in the configuration example for realizing the first embodiment of the present invention. In the figure, a general cluster refers to a cluster extracted by the cluster extraction unit 15 and the cluster analysis unit 16, and a special cluster refers to a set of pixels that do not belong to any cluster among pixels of image data.
[0064]
Here, for the general cluster, color conversion into the color reproduction range is performed so that the color discrimination can be maintained for the flat portion of the input image. Basically, color conversion is not performed within the color reproduction range, and when a color outside the color reproduction range is included, a preset initial profile is used. Further, the gradation portion is subjected to color conversion within the color reproduction range so that the gradation can be maintained. If it is within the color reproduction range, color conversion is not performed. If a color outside the color reproduction range is included, a color conversion is performed by creating a profile depending on the color distribution in the cluster. For special clusters, color conversion is performed using a preset initial profile.
[0065]
FIG. 8 is a flowchart showing an example of the operation of the cluster-specific color conversion unit 17 in the configuration example for realizing the first embodiment of the present invention. First, in S131, an unprocessed cluster c_i Select one. Subsequent processing is the cluster c selected here._i Is the target cluster.
[0066]
In S132, the target cluster c_iIt is checked whether the color distribution of is flat. This is determined based on the color distribution obtained by the cluster analysis unit 16. Target cluster c_iIf the color distribution is flat, the processes of S133, S134, and S135 are performed. If it is not flat, i.e., gradation, the processes of S136, S137, and S138 are performed.
[0067]
Target cluster c_iIn step S133, a list of clusters having a flat color distribution is created. Here, the target cluster c_i To the flat list. In this example, for the sake of efficiency, color conversion is performed using a single profile for a cluster having a flat color distribution and a special cluster. The process of S133 is a preprocess for this purpose.
[0068]
In S134, the target cluster c_i It is checked whether or not includes pixels outside the color reproduction range. If so, the target cluster c is obtained in S135._i Representative color M_i The color conversion of representative colors is performed by applying an initial profile set in advance. The result of this color conversion is recorded in a flat list.
[0069]
The initial profile in this example is composed of a set of conversion vectors that move only the colors outside the color reproduction range in the saturation direction without changing the lightness and hue and paste them to the outer edge of the color reproduction range. . No conversion is performed for colors within the color reproduction range. FIG. 9 shows a cluster c having a flat color distribution._i Representative color M_i Is an explanatory diagram when color conversion is performed based on an initial profile. In this figure, for the sake of simplicity, a plane represented by saturation and hue is shown. In the figure, the solid line is the outer edge of the color reproduction range, and the black circle is a flat cluster c._i Representative color M_i The solid arrow is the representative color M_i Transformation vector u corresponding to_i It is. As shown in FIG. 9, for example, the representative color M outside the color reproduction range.₂, M_Five, M₆Is the transformation vector u₂, u_Five, u₆Is converted to the outer edge of the color reproduction range. At this time, the hue and brightness are maintained, and only the saturation is changed. The representative color M within the color reproduction range₁, M_Three, M_Four, M₇, M₈Does not convert.
[0070]
Target cluster c_iIf the color distribution of the cluster is gradation, the cluster c in S136._i The area on the image plane occupied by is added to the gradient mask. Eventually, a mask corresponding to an area occupied by all clusters whose color distribution is gradation is created. Since gradation conversion is performed for each cluster and color conversion for other portions is performed later, it is created in order to obtain an area to be subjected to color conversion for other portions.
[0071]
In S137, the target cluster c_i It is checked whether or not includes pixels outside the color reproduction range. If it is not included, no conversion into the color conversion range is required, and the process proceeds to S139 without doing anything. If it is included, a cluster c having a gradation color distribution is obtained in S138._iIs converted to a color reproduction range that retains gradation. Here, the target cluster c_i The vector v having the largest distance between the pixel outside the color reproduction range belonging to and the outer edge of the color reproduction range_i For this vector v_i Is used as a conversion vector to convert the color of a pixel whose color distribution including a pixel outside the color reproduction range belongs to a gradation cluster.
[0072]
FIG. 10 shows a cluster c in which the color distribution including colors outside the color reproduction range is gradation._i It is explanatory drawing in the case of color-converting the color of the pixel which belongs to to the color reproduction range. In this figure, for the sake of simplicity, a plane represented by saturation and hue is shown. In the figure, the solid line is the outer edge of the color reproduction range, and the black circle is the cluster c_i Is a color of a pixel belonging to, and a solid arrow indicates a vector v_i It is. The dotted arrow indicates cluster c._i The color of a pixel belonging to the vector v_i The black square indicates the color of the pixel after color conversion. Vector v with the largest distance from the outer edge of the color reproduction range_i By performing color conversion using_iAll the colors are within the color reproduction range. At this time, cluster c_iSince the relative change amount between each color is stored as it is, the gradation is maintained.
[0073]
In this example, cluster c_i The color of all pixels belonging to the vector v_i However, the present invention is not limited to this, and any method can be used as long as the color of the pixel belonging to the target cluster can be converted into the color reproduction range while maintaining the gradation in the cluster. May be. For example, in order to minimize the influence of color conversion on other colors, the cluster c_i The color distribution of all pixels belonging to the vector v_i The color may be converted within the color reproduction range after being reduced according to the size of the image.
[0074]
In S139, it is checked whether or not an unprocessed cluster still remains. If it remains, the process returns to S131 and the process is repeated. If not, the processes of S140 and S141 are performed. Immediately before performing the process of S140, a flat list that is a list of clusters having a flat color distribution is completed. Note that the representative color after color conversion by the process of S135 is also recorded in this flat list, and the representative color M outside the color reproduction range is recorded._i Is M_iShall be converted to '. Representative color M within the color reproduction range_i Is the representative color M after the conversion_iEqual to '. In addition, for a cluster having a gradation color distribution, color conversion has already been completed by the processing of S138.
[0075]
FIG. 11 is an explanatory diagram of an example of a flat list. This figure is created based on the example shown in FIG. Therefore, the cluster c in which the representative color exists within the color reproduction range₁, c_Three, c_Four, c₇, c₈ The representative color is equal to the converted representative color. In addition, cluster c₂, C_Five, C₆Since the representative color is outside the color reproduction range, the representative color M is outside the color reproduction range.₂, M_Five, M₆ Representative color M converted from₂', M_Five', M₆'Is the color of the outer edge of the color reproduction range.
[0076]
In S140, referring to the flat list, the representative color M after color conversion is displayed._iIs scanned to determine whether there is a cluster having a representative color similar to the representative color after color conversion of the cluster outside the color reproduction range. If this is present, there is a possibility that the distinguishability from other colors may be deteriorated by color conversion. In order to prevent such deterioration of color discrimination, the initial profile is corrected. Here, the representative color M after color conversion of the cluster outside the color reproduction range._iThe difference between the hue of 'and the predetermined threshold Th₉ Extract smaller clusters as likely to be similar.
[0077]
FIG. 12 is an explanatory diagram showing an example of a method for examining whether or not there is a cluster having a representative color similar to the representative color after color conversion of a cluster outside the color reproduction range. In this figure, the representative color M of the cluster outside the color reproduction range.₂ About its hue H₂ The difference in hue with Th₉ Search for clusters with smaller representative colors. In this example, the representative color M₁ And M_Three Is Hue H₂The difference in hue with Th₉Smaller than. Therefore, representative color M₂The color of a cluster with₁And M_ThreeThis representative color M may be similar to a cluster having₁, M_ThreeIs extracted.
[0078]
Then, the color difference between the representative color extracted as having the possibility of being similar and the converted representative color is calculated, and this color difference is determined by a predetermined threshold Th._TenIf it is smaller, the color discriminability is determined to be reduced. In the present invention, the present invention is not limited to this method, and any method may be used as long as it can extract a representative color having a similar color to the extent that the color discrimination is reduced.
[0079]
If it is determined that the color discrimination is deteriorated, the conversion vector in the initial profile corresponding to the representative color of the cluster outside the color reproduction range that causes the color change or the color in the vicinity thereof is not reduced. Correct it. Here, the representative color M after color conversion of the cluster outside the color reproduction range._iOne or more representative colors M extracted as possibly similar to '_j And the color difference is Th_TenThe above and the representative color M after color conversion_iM without changing the hue of '_iThe representative color M with the smallest color difference from '_i"Calculate M_i To M_iIt is corrected by replacing it with a conversion vector to ".
[0080]
FIG. 13 shows the representative color M when it is determined that the color discrimination is deteriorated._iIt is explanatory drawing of an example of the process which calculates ".₂ Is a color outside the color reproduction range, M₂'Is a point on the outer edge of the color reproduction range, M₁ And M_Three Is M₂Color determined to be similar to 'M,_Three Is a color that has been determined to further deteriorate the discrimination. FIG. 13 shows a three-dimensional space, where the solid line indicates the outer edge of the color reproduction range and the hue is H.₂The line of intersection with a plane in the space is shown. Further, the broken line indicates each color M.₁, M_ThreeTo radius Th_TenThe spherical surface is shown. M to find here₂"Is the outer edge of color reproduction range and color M_Three Radius Th_TenIs the point of intersection with the spherical surface. Thus, M₂ To M₂"Transform vector u to₂'Can be calculated and replaced with the corresponding transformation vector of the initial profile.
[0081]
In this way, the initial profile is corrected to prevent the color discrimination from being deteriorated, and a profile for performing color conversion of portions other than the cluster having the gradation color distribution is created.
[0082]
In S141, color conversion is performed by applying the profile created in S140 to pixels other than the gradation mask generated in S136.
[0083]
As described above, the color conversion processing performed by the cluster-specific color conversion unit 17 is performed in units of clusters. The processing efficiency can be improved by targeting the cluster limited to the important color as in the above example, and the obtained result also suppresses the change in the reproducibility of the entire image by color conversion.
[0084]
Returning to FIG. 2, the output-side color conversion unit 18 can output the color system image data used in the internal processing converted into the color within the color reproduction range by the cluster-specific color conversion unit 17. Convert to color system image data. This conversion can be performed by an existing method. In this example, L^* C^* Because the H ° color system was used for internal processing, L^* C^* The H ° image data is converted into, for example, YMCK color system image data that can be output by the output device. The image data output device 19 outputs image data that can be output by the output device, generated by the output side color conversion unit 18.
[0085]
In this way, even when image data that includes colors that cannot be reproduced by the output device is input, color gradation is preserved in the gradation portion, and distinctiveness is preserved in the flat portion, so that reproducible colors can be obtained. Good reproducibility can be obtained by converting the data into and outputting. At this time, by particularly performing such conversion for the important colors, it is possible to improve the reproducibility of the entire image as well as the processing efficiency.
[0086]
FIG. 14 is a conceptual block diagram showing a second embodiment for realizing the image processing apparatus and the image processing method of the present invention. In the figure, the same parts as those in FIG. Reference numeral 6 denotes an important area creation unit, 7 denotes a group creation unit, and 8 denotes a color conversion unit. The important area creating unit 6 receives an instruction for an area in the image where it is desired not to change the color as much as possible, and creates an important area in accordance with the instruction. Based on the important area created by the important area creating section 6 and the important color extracted by the image data analyzing section 2 from the image data input by the image data input section 1, the group creating section 7 determines the color and its distribution. Create similar groups. For each group generated by the group creation unit 3, the color conversion unit 8 includes colors in the group and their distribution, important regions created by the important region creation unit 6, important colors extracted by the image data analysis unit 2, A color conversion vector is determined based on the color reproduction range, and color conversion of the image data input by the image data input unit 1 is performed.
[0087]
The output target image data input by the image data input unit 1 is quantized and supplied to the image data analysis unit 2, group creation unit 3, and color conversion unit 4. On the other hand, an instruction regarding an area in the image in which the color is not desired to be changed is input, and an important area is created in the important area creating unit 6. The image data analysis unit 2 extracts one or more colors determined to be important from the given image data as important colors, and creates a group based on the important colors and the important regions created by the important region creation unit 6. The unit 7 generates groups having similar colors and distributions from the given image data. The color conversion unit 8 creates a color conversion vector based on the generated group and the color distribution in the group, the important region, the important color, and the color reproduction range of the output device, and applies this to the given image data. To perform color conversion. The image data after color conversion is output from the image data output unit 5 to the output device.
[0088]
FIG. 15 is a block diagram showing a configuration example for realizing the second embodiment of the present invention. In the figure, parts similar to those in FIG. Reference numeral 21 denotes an important region creation unit, 22 denotes a cluster extraction unit, and 23 denotes a cluster-specific color conversion unit. The important area creating unit 21 receives an instruction of an area in the input image that does not change greatly as a whole by color conversion, but preferably does not change the color as much as possible, and uses this area as an important area as the temporary data storage device 13. To record. Here, an input image is displayed, a closed region is designated using a pointing device such as a mouse, and this closed region is set as an important region. If there is no particular need, these important areas need not be specified. The present invention is not limited to the method described here, and any method may be used as long as it can obtain a region where it is preferable that the color is not changed as much as possible. For example, when inputting an image, a binary image representing these areas may be read. In this example, it is assumed that the created important region contains exactly the target region. However, by specifying the target region roughly and investigating the vicinity of the boundary, it is possible to obtain the correct region. It can also be configured.
[0089]
The cluster extraction unit 22 extracts a cluster which is a set of pixels having similar characteristics based on the important region created by the important region creation unit 21 and the important color extracted by the important color extraction unit 14. For example, a cluster can be efficiently extracted by forming a cluster only from pixels included in an important area and pixels having an important color or a color close to the important color. In this example, for colors other than the important areas that are significantly different from the important colors, the color reproducibility of the entire image is assumed not to change significantly by color conversion, and special processing for preserving color discrimination and gradation is performed. Absent.
[0090]
In this example, L^* C^* Clusters are extracted by clustering according to a rearrangement method on a five-dimensional feature space of brightness, saturation, hue, and X and Y coordinates on the image plane in the H color space. As an example of the operation of the cluster extraction unit 22, processing may be performed according to the flowchart shown in FIG. 4 as in the first embodiment described above. In the second embodiment, when the initial cluster is generated in S101 of FIG. 4, the color difference from the important color is the threshold Th.₂ Not only the smaller pixels but also the important area itself is the initial cluster. In this way, since the pixels included in the important region also belong to the extracted cluster, the color conversion that retains the gradation and color discrimination of the important region through the cluster analysis unit 16 and the cluster-specific color conversion unit 23. Can be realized.
[0091]
In this example, the importance color and the importance area are regarded as being equivalent, and such processing is performed. However, if the importance color and the importance area are differentiated, for example, the importance of the importance area is emphasized. It is also possible to perform the processing of the cluster extraction unit 15 for pixels in an image other than the important region, and add one important region as one cluster thereto. In addition, an attribute of an important area may be added to the cluster so that the important area is not unnecessarily divided, integrated with other areas, or further deleted. Furthermore, a level may be provided in the important area, and the processing may be divided according to the level.
[0092]
The cluster-specific color conversion unit 23 can be configured in the same manner as the cluster-specific color conversion unit 17 in the first embodiment described above, but here, for each cluster, its importance I_i And an example in which the color of the pixels of the cluster with high importance is not converted as much as possible will be described.
[0093]
The importance of the cluster is cluster c_i Total number of pixels belonging to Nc_i And the number of pixels in the important area, the color difference between the important color and the representative color, and the like. In this example, cluster c_i Total number of pixels belonging to Nc_i Is used as the cluster importance as it is. Therefore, cluster c_i Total number of pixels belonging to Nc_i The more it is, the more important the cluster.
[0094]
The processing in the cluster-specific color conversion unit 23 is the same as that of the cluster-specific color conversion unit 17 in the first embodiment described above, and the processing may be performed according to the flowchart shown in FIG. However, in this second embodiment, importance is used, and when color conversion having a flat color distribution having a representative color outside the color reproduction range is converted into a color reproduction range, the color discrimination is reduced. The method for correcting the conversion vector after the determination is changed.
[0095]
In the first embodiment described above, as described with reference to FIG. 13, when it is determined that the color discriminability is deteriorated, a cluster having a flat color distribution having a representative color that exists outside the color reproduction range is uniform. The conversion vector to convert was corrected. In the second embodiment, the cluster importance level is used, and the conversion vector of the cluster with low importance level is corrected. Accordingly, when a cluster having a representative color that exists outside the color reproduction range and having a flat color distribution is less important than a cluster within the color reproduction range in which the color discrimination performance is reduced, it is shown in FIG. As described above, the conversion vector that converts a cluster with a flat color distribution with a representative color that exists outside the color reproduction range is corrected. The representative color of the cluster is converted to a color that does not deteriorate color discrimination.
[0096]
FIG. 16 is an explanatory diagram of an example of a process for calculating a conversion vector for converting the representative colors of the clusters in the color reproduction range so that the color discrimination is not deteriorated. In the figure, M₂ Is a color outside the color reproduction range, M₂'Is a point on the outer edge of the color reproduction range, M₁ And M_Three Is M₂Color determined to be similar to 'M,_Three Is a color that has been determined to further degrade color discrimination. FIG. 16 shows a three-dimensional space. The solid line indicates the outer edge of the color reproduction range and the hue is H.₂The line of intersection with a plane in the space is shown. Further, the broken line indicates the radius Th._TenThe spherical surface is shown.
[0097]
Here, representative color M₂Of the cluster with, and the representative color M_ThreeWhen comparing the importance of clusters having₂Suppose that the cluster with is more important. In this case, as shown in FIG.₂Transformation vector u₂Instead of changing the color M, as shown in FIG._ThreeMove. For example, color M_Three Radius Th_TenMove the ball of M,₂'At the contact point and M_Three Color M with the smallest color difference from_Three'Is calculated. And M_Three M_ThreeConversion vector u to convert to '_Three Is replaced with the corresponding conversion vector of the initial profile by this conversion vector. By configuring in this way, it is possible to reduce the width of change for a color that causes a greater change in reproducibility by color conversion in the image.
[0098]
【The invention's effect】
As is clear from the above description, according to the present invention, the color distribution is grouped according to the input image data, and the conversion vector is determined so as to be within the color reproduction range for each group. It is possible to prevent the color discrimination and gradation from being lowered without greatly changing the color of the color. Furthermore, by performing grouping limited to important colors or colors similar to important colors and important colors, color conversion into the color reproduction range can be performed efficiently. In addition, by introducing the degree of change in the reproducibility of the entire image by changing the color as an importance level, color conversion within the color reproduction range can be performed so that color reproduction faithful to the input image can be performed. There is an effect that it can be realized. In addition, even in a relatively small portion of an image, gradation such as gradation can be maintained by designating the region as an important region.
[Brief description of the drawings]
FIG. 1 is a conceptual block diagram showing a first embodiment for realizing an image processing apparatus and an image processing method of the present invention.
FIG. 2 is a block diagram showing a configuration example for realizing the first embodiment of the present invention.
[Figure 3] L^*C^*It is explanatory drawing of the unit space created by dividing the H ° color space.
FIG. 4 is a flowchart illustrating an example of the operation of the cluster extraction unit 15 in the configuration example that implements the first exemplary embodiment of the present invention.
FIG. 5 is a flowchart illustrating an example of an operation of cluster color distribution recognition processing of a cluster analysis unit 16 in a configuration example that implements the first exemplary embodiment of the present invention;
FIG. 6 is a flowchart showing an example of cluster integration and deletion processing operations of the cluster analysis unit 16 in the configuration example for realizing the first exemplary embodiment of the present invention;
FIG. 7 is an explanatory diagram illustrating an example of correspondence between color conversion performed by a cluster-specific color conversion unit and cluster color distribution in a configuration example that implements the first exemplary embodiment of the present invention;
FIG. 8 is a flowchart illustrating an example of the operation of the cluster-specific color conversion unit 17 in the configuration example that implements the first exemplary embodiment of the present invention.
FIG. 9 is a cluster c with a flat color distribution._i Representative color M_i Is an explanatory diagram when color conversion is performed based on an initial profile.
FIG. 10 is a cluster c in which a color distribution including a color outside the color reproduction range is gradation._i It is explanatory drawing in the case of color-converting the color of the pixel which belongs to to the color reproduction range.
FIG. 11 is an explanatory diagram of an example of a flat list.
FIG. 12 is an explanatory diagram illustrating an example of a method for examining whether or not there is a cluster having a representative color similar to the representative color after color conversion of a cluster outside the color reproduction range.
FIG. 13 shows a representative color M when it is determined that the color discrimination is deteriorated._iFIG. 11 is an explanatory diagram of an example of a process for calculating “.
FIG. 14 is a conceptual block diagram showing a second embodiment for realizing the image processing apparatus and the image processing method of the present invention.
FIG. 15 is a block diagram illustrating a configuration example for realizing a second exemplary embodiment of the present invention.
FIG. 16 is an explanatory diagram illustrating an example of a process for calculating a conversion vector for converting the representative colors of the clusters in the color reproduction range so that the color discrimination characteristics are not deteriorated.
[Explanation of symbols]
DESCRIPTION OF SYMBOLS 1 ... Image data input part, 2 ... Image data analysis part, 3 ... Group creation part, 4 ... Color conversion part, 5 ... Image data output part, 6 ... Important area creation part, 7 ... Group creation part, 8 ... Color conversion 11: Image data input device, 12 ... Input side color conversion unit, 13 ... Temporary data storage device, 14 ... Important color extraction unit, 15 ... Cluster extraction unit, 16 ... Cluster analysis unit, 17 ... Color conversion unit for each cluster , 18 ... output side color conversion unit, 19 ... image data output device, 21 ... important area creation unit, 22 ... cluster extraction unit, 23 ... color conversion unit for each cluster.

Claims (16)

  Image data input means for inputting image data, image data analysis means for extracting one or more important colors from the image data input by the image data input means, and the important data extracted by the image data analysis means A group creating means for recognizing the color of the image data and the distribution on the image plane based on the color and the color similar to the important color and generating a group for each different color and the distribution on the image plane; Group analysis means for recognizing the color distribution for each group generated by the group creation means and classifying the group into gradation or flat, and the important color extracted by the image data analysis means and the color reproduction range of the output device An image processing system comprising color conversion means for performing color conversion according to the classification by the group analysis means based on Apparatus.   One or more of image data input means for inputting image data, area instruction means for indicating an important area in the image data input by the image data input means, and one or more of the image data input by the image data input means Image data analysis means for extracting the important color of the image data, the image data based on the important color extracted by the image data analysis means, a color similar to the important color, and the important area designated by the area designation means Creating a group for each color and its distribution on the image plane and generating a group for each different color and its distribution on the image plane, and recognizing the color distribution for each group generated by the group creation means Group analysis means for classifying the group into gradation or flat, and important data extracted by the image data analysis means The image processing apparatus characterized by comprising a color converter for performing color conversion in accordance with the classification by the group analysis unit based on the color reproduction range of an output device.   The image data analysis unit calculates a frequency in the color space of the image data input by the image data input unit, and extracts a color having a frequency or a ratio of the frequency to the whole of a threshold value or more as an important color. The image processing apparatus according to claim 1, wherein the image processing apparatus is an image processing apparatus.   The image data analysis means passes the important color to the group creation means when the image data input by the image data input means includes a color outside the color reproduction range in the output device. The image processing apparatus according to claim 3.   The image processing apparatus according to claim 1, wherein the group creating unit creates a group by clustering based on an important color extracted by the image data analyzing unit.   The image processing apparatus according to claim 5, wherein the group creating unit changes a clustering threshold value or a parameter between a cluster including an important color extracted by the image data analyzing unit and a cluster not including the important color.   The image processing apparatus according to claim 1, wherein the color conversion unit determines a color conversion parameter so as not to change the important color extracted by the image data analysis unit as much as possible, and performs color conversion.   The color conversion means, if there is an important color extracted by the image data analysis means and an important area designated by the area instruction means, color conversion parameters so as not to change the color and gradation in the area as much as possible. The image processing apparatus according to claim 2, wherein color conversion is performed.   One or more important colors are extracted from the input image data by image data analysis means, and based on the extracted important colors and colors similar to the important colors, the color of the image data and its image plane Recognizing the distribution and generating a group for each different color and its distribution on the image plane by the group creating means, recognizing the generated color distribution for each group and making the group gradation or flat by the group analyzing means An image processing method comprising: classifying and performing color conversion according to the classification based on the important color and a color reproduction range of an output device by a color conversion unit.   An instruction about an important area in the image data input by the image data input means is received, while one or more important colors are extracted from the input image data by the image data analysis means, and the extracted important data Recognize the color of the image data and its distribution on the image plane based on the color, the color similar to the important color and the important area, and create a group for each different color and its distribution on the image plane The group is classified into gradation or flat by the group analysis unit by recognizing the generated color distribution for each group, and according to the classification based on the important color and the color reproduction range of the output device An image processing method characterized in that color conversion is performed by color conversion means.   When extracting the important color, a frequency in the color space of the image data is calculated, and a color whose frequency or a ratio of the frequency to the whole is equal to or more than a threshold is extracted as an important color. Item 11. The image processing method according to Item 9 or 10.   When extracting the important color, the color outside the color reproduction range in the output device of the input image data and the color estimated to be converted outside the color reproduction range are extracted as the important color. The image processing method according to claim 9 or 10, wherein:   The image processing method according to claim 9 or 10, wherein when generating the group, the group is generated by clustering based on the important color.   14. The image processing method according to claim 13, wherein the clustering is performed by changing a clustering threshold value or a parameter between a cluster including the important color and a cluster not including the important color.   11. The image processing method according to claim 9, wherein when performing the color conversion, color conversion parameters are determined so as not to change the important color as much as possible, and the color conversion is performed.   When performing the color conversion, if the important color and the designated important area exist, color conversion parameters are determined so as not to change the color and gradation in the area as much as possible, and the color conversion is performed. The image processing method according to claim 10.

Images