JP4248947B2 - Image color reduction device, image color reduction method, and image color reduction program - Google Patents

Description

の関係式を用いてＲＧＢ表色系からＹＣ表色系への変換をおこなう。ここで、Ｙは、ＹＣ表色系における輝度の値であり、Ｃ１’およびＣ２’は、ＹＣ表色系における２つの色差の値であり、Ｒ、ＧおよびＢは、それぞれＲＧＢ表色系の赤、緑および青の階調値を表す。
【００２８】
さらに、色空間変換部１２は、色差Ｃ１’およびＣ２’を１バイトのデータ量で表すことができるように、
Ｃ１＝ｍａｘ（０，ｍｉｎ（２５５，Ｃ１’＋１２８）） ．．．．．．（４）
Ｃ２＝ｍａｘ（０，ｍｉｎ（２５５，Ｃ２’＋１２８）） ．．．．．．（５）
の関係式を用いて色差Ｃ１’およびＣ２’の値を変換する。ここで、ｍａｘ（ａ，ｂ）は、ａおよびｂのうち大きい方の値を意味し、ｍｉｎ（ａ，ｂ）は、ａおよびｂのうち小さい方の値を意味する。また、変換された色差Ｃ１およびＣ２は、０から２５５までの値をとる。
【００２９】
図３に、ＲＧＢ表色系からＹＣ表色系に変換して得られた画像データの色空間の一例を示す。色差Ｃ１およびＣ２を横軸に、輝度Ｙを縦軸に設定し、色の分布が点の集まりとして表されている。また、図３には示されていないが、色空間内の各点は座標値だけでなく、色分布の度数の情報をさらに保持している。
【００３０】
図１の説明に戻ると、２次元ヒストグラム極大値探索部１３は、色空間のＣ１−Ｃ２平面に係る２次元ヒストグラムを、色空間変換部１２により変換された色の分布に基づいて作成し、作成された２次元ヒストグラムの極大値を探索する探索部である。
【００３１】
具体的には、まず、画像のエッジ部分の色が代表色として選定されないようにするために、エッジ強度を考慮したヒストグラムを作成する処理をおこなう。これは、たとえば、白の領域内に赤の領域がある帳票をスキャンした際に、スキャンして得られた画像の赤の領域のエッジの部分の色が、もとの帳票にはないピンクの領域となる場合があり、そのような場合に、赤ではなくピンクを代表色として選択してしまうといったことを抑制するためである。
【００３２】
このエッジ強度は、ソーベルフィルタを画像の各画素の輝度Ｙの値に対して適用することにより算出する。図４は、各画素の重み付けに用いるソーベルフィルタのオペレータを示す図である。図４（ａ）は、縦方向のエッジを検出するためのオペレータであり、図４（ｂ）は、横方向のエッジを検出するためのオペレータである。また、図４（ｃ）は、注目している画素とその近傍８画素の輝度Ｙの階調値を表す。
【００３３】
ソーベルフィルタを適用することにより、エッジ強度ｗ１は、

により算出することができる。ここで｜｜は、値の絶対値を取ることを意味する。エッジ強度ｗ１は、縦方向および横方向のエッジを検出するためのオペレータを適用した値の絶対値の和であるので、縦横両方の方向のエッジを考慮することができる。
【００３４】
そして、算出されたエッジ強度ｗ１を基にして、画像の各画素のヒストグラム加算値ｗを、
ｗ＝ｍａｘ（０，１−ｗ１／ｅｄｇｅ） ．．．．．．（９）
により算出する。ここで、ｍａｘ（ａ，ｂ）は、ａおよびｂのうち大きい方の値を意味し、ｅｄｇｅは、１以上の所定の定数である。
【００３５】
続いて、色空間内の座標点（Ｃ１，Ｃ２，Ｙ）に対応するすべての画素のヒストグラム加算値ｗを足し合わせ、その足し合わされた値を座標点（Ｃ１，Ｃ２，Ｙ）に対する度数とすることにより３次元ヒストグラムｈ（Ｃ１，Ｃ２，Ｙ）を作成する。
【００３６】
その後、色空間のＣ１−Ｃ２平面に係る２次元ヒストグラムＨ₁（Ｃ１，Ｃ２）を、
Ｈ₁（Ｃ１，Ｃ２）＝Σｈ（Ｃ１，Ｃ２，Ｙ） ．．．．．．（１０）
により作成する。ここで、Σは、Ｃ１およびＣ２で指定される色差の座標に対応するすべての点の度数ｈ（Ｃ１，Ｃ２，Ｙ）の和をとることを意味する。
【００３７】
ここで、式（９）において、ｅｄｇｅの値が小さくなるほど、エッジ強度ｗ１の値がヒストグラム加算値ｗに強く反映される。逆に、ｅｄｇｅの値が大きくなるほど、エッジ強度ｗ１の値がヒストグラム加算値ｗに反映されなくなる。したがって、ｅｄｇｅの値を適切に設定することにより、画像のエッジ部分の色を代表色として選定しないようにする程度を調整することができる。
【００３８】
Ｃ１−Ｃ２平面に係る２次元ヒストグラムＨ₁（Ｃ１，Ｃ２）が作成された後、２次元ヒストグラム極大値探索部１３は、２次元ヒストグラムＨ₁（Ｃ１，Ｃ２）の値を調べ、注目している色差の点（Ｃ１，Ｃ２）の度数がその色差の点をＣ１−Ｃ２平面上で取り囲む近傍８点の度数よりも大きく、かつ、その注目している色差の点の度数が全体の度数（すべての点の度数の和）の０．０００１％以上である場合に、その度数を示す色差の点を、度数の極大値を示す極大点として抽出する処理をおこなう。このように、度数に対するしきい値を設定し、度数がある程度大きいもののみを、極大点として選択する。
【００３９】
図５は、Ｃ１−Ｃ２平面における２次元ヒストグラムの一例を示す図である。同図に示すように、この２次元ヒストグラムは、色差Ｃ１およびＣ２の値が横軸に、エッジ強度ｗ１を考慮した度数の値Ｈ₁（Ｃ１，Ｃ２）が縦軸に設定されている。また、図５には、米印で示される１１個の極大点の位置が示されている。
【００４０】
図１の説明に戻ると、２次元ヒストグラム極大値探索部１３は、さらに、得られたＣ１−Ｃ２平面に係る２次元ヒストグラムの平滑化処理をおこなう。これは、２次元ヒストグラムを基にして抽出した極大点の数を制御するためである。
【００４１】
この平滑化処理は、Ｃ１−Ｃ２平面に係る２次元ヒストグラムに平滑化フィルタを適用することによりおこなう。図６は、Ｃ１−Ｃ２平面に係る２次元ヒストグラムに適用される平滑化フィルタのオペレータを示す図である。図６（ａ）は、注目している色差の点（Ｃ１，Ｃ２）とその近傍８点の色差の点に適用されるオペレータであり、図６（ｂ）は、注目している色差の点（Ｃ１，Ｃ２）とその近傍８点の色差の点に対応する度数の値を表す。
【００４２】
平滑化フィルタを適用することにより、色差の点（Ｃ１，Ｃ２）に対応する２次元ヒストグラムＨ₂（Ｃ１，Ｃ２）は、

により算出される。平滑化処理をおこなった後、２次元ヒストグラム極大値探索部１３は、２次元ヒストグラムＨ₂（Ｃ１，Ｃ２）の極大点の探索をおこなう。
【００４３】
図１の説明に戻ると、１次元ヒストグラム極大値探索部１４は、２次元ヒストグラム極大値探索部１３により探索された極大点を通り、ＹＣ表色系の色空間のＣ１−Ｃ２平面に垂直な直線に係る１次元ヒストグラムを作成し、作成された１次元ヒストグラムの極大値を探索する探索部である。
【００４４】
図７は、Ｃ１−Ｃ２平面に係る２次元ヒストグラムの極大点を通るＣ１−Ｃ２平面に垂直な直線の一例を示す図である。同図に示すように、これらの直線は、２次元ヒストグラム極大値探索部１３により探索された極大点を通り、Ｃ１−Ｃ２平面に垂直に伸びている。
【００４５】
図８は、１次元ヒストグラムの作成方法を説明する概念図である。同図に示すように、１次元ヒストグラムを作成する場合には、色空間内において、Ｃ１−Ｃ２平面に垂直で極大点を通る直線を考え、その直線上の各点Ｋ（Ｃ１，Ｃ２，ｋ）から所定の距離Ｄ内にある点の度数の和Ｈ₃（Ｃ１，Ｃ２，ｋ）を、
Ｈ₃（Ｃ１，Ｃ２，ｋ）＝Σ｛ｈ（Ｃ１，Ｃ２，ｋ）×ｗ２｝ ．．．（１２）
により算出し、この値を点Ｋにおける１次元ヒストグラムの度数とする。ここで、Ｄ≦ｋ≦２５５−Ｄである。
【００４６】
また、ｈ（Ｃ１，Ｃ２，ｋ）は、式（９）のヒストグラム加算値により作成された３次元ヒストグラムであり、ｗ２は、ｈ（Ｃ１，Ｃ２，ｋ）に対して付与される重みである。この重みｗ２は、
ｗ２＝ｅｘｐ（−Ｌ²／５０−Ｍ²／２−Ｎ²／２） ．．．．．．（１３）
により算出される。ここで、Ｌ、ＭおよびＮは、それぞれ所定の距離Ｄ内にある点と点Ｋ（Ｃ１，Ｃ２，ｋ）と間の、Ｙ軸方向、Ｃ１軸方向、Ｃ２軸方向それぞれの方向への距離である。重みｗ２は、点Ｋ（Ｃ１，Ｃ２，ｋ）からの距離が大きくなるほど小さくなる。Σは、所定の距離Ｄ内にあるすべての点の度数ｈ（Ｃ１，Ｃ２，ｋ）に重みｗ２をかけた値の和をとることを意味する。
【００４７】
図９は、１次元ヒストグラムにおける極大点の概念を示す図である。同図に示すように、この例では、上記に示した方法で１次元ヒストグラムを作成した結果、２つの極大点が直線上にあることがわかる。この極大点を抽出することにより、画像の色を代表する代表色の候補となる色を抽出することができる。
【００４８】
具体的には、各極大点に対応する線分上の１次元ヒストグラムにおいて、度数が線分上の両隣の点の度数よりも大きく、かつ、その度数がその線分上における各点の度数の合計の０．００１％よりも大きい場合に、その点を極大点として抽出する処理をおこなう。
【００４９】
上記の２次元ヒストグラムの平滑化処理、平滑化された２次元ヒストグラムにおける極大点の探索処理、および、１次元ヒストグラムにおける極大点の探索処理は、１次元ヒストグラム探索部１４により所定の条件が満足されたと判定されるまで、繰り返しおこなわれる。
【００５０】
具体的には、１次元ヒストグラム探索部１４は、１つ前の繰り返しの際に探索された各極大点に対応する色を記憶しておき、それらと新しく探索された各極大点に対応する色との間の色空間内でのユークリッド距離を算出し、ユークリッド距離の最大値があらかじめ設定されたしきい値よりも大きいかどうかを調べる処理をおこなう。そして、ユークリッド距離の最大値がしきい値よりも大きくなった場合に、１つ前の繰り返しの際に探索された極大点を代表色の候補として採用し、繰り返し処理を終了する。
【００５１】
また、新しく探索された極大点の数が、あらかじめ設定された極大点の最小数を下回った場合、あるいは、繰り返しの回数が所定の回数を越えた場合には、１つ前の繰り返しの際に探索された極大点を代表色の候補として採用し、繰り返し処理を終了する。そのほかの場合には、極大点の統合がまだ可能であると判定し、繰り返し処理を継続する。
【００５２】
図１０は、平滑化処理による極大点の数および代表色候補の色数の変化を示す図である。同図に示すように、この例では、１回目の平滑化においては、２次元ヒストグラムにおける極大点は１１個存在し、１次元ヒストグラムにおける極大点に対応する色、すなわち、代表色の候補となる色は１７色存在する。ところが、平滑化回数が多くなるにつれ、似通った色差の特徴を有する極大点どうしが統合され、極大点の数および代表色の候補となる色の数が少なくなる。
【００５３】
図１１は、平滑化処理がなされたＣ１−Ｃ２平面における２次元ヒストグラムの一例を示す図である。この図では、図５に示した１１個の極大点を持つ２次元ヒストグラムに対して平滑化処理がなされ、極大点の数が５個に減少した様子を示している。
【００５４】
図１の説明に戻ると、色置換部１５は、１次元ヒストグラム極大値探索部１４により探索された１次元ヒストグラムの極大値に対応する色を代表色として選定し、選定された代表色で画像に含まれる色を置換する置換部である。
【００５５】
具体的には、探索された極大点に対応する色（代表色の候補色）のすべての組に対して、評価値Ｅを、
Ｅ＝（候補色ｉおよび候補色ｊ間のユークリッド距離）
＋α×ｍｉｎ（全候補色の度数の和に対する候補色ｉの度数の比率，
全候補色の度数の和に対する候補色ｊの度数の比率）
．．．．．．（１４）
により算出する。ここで、（候補色ｉおよび候補色ｊ間のユークリッド距離）は、色空間（Ｃ１，Ｃ２，Ｙ）における２色間のユークリッド距離であり、候補色の度数の値は、１次元ヒストグラムにおける度数の値であり、ｍｉｎ（ａ，ｂ）は、ａおよびｂのうち小さい方の値を意味し、αは、所定の正の定数である。
【００５６】
この評価値Ｅが、あらかじめ定められたしきい値よりも小さい場合には、全候補色の度数の和に対する度数の比率が小さい候補色が削除され、度数の比率が大きい候補色に統合される。ここでは、評価値Ｅを算出する式（１４）において、候補色の度数の比率を考慮しているが、これは、度数の値が大きい候補色が、その値が大きいにもかかわらず削除され、代表色として選ばれなくなることを防ぐためである。
【００５７】
色置換部１５は、このようにして得られた候補色を代表色として選定し、ＹＣ色空間内において、もとの画像の画素の色との間のユークリッド距離が最も小さくなる代表色で、もとの画像の画素の色を置換する置換部である。その際、代表色に置換した画素数を代表色ごとにカウントし、全代表色のカウント数に対するそのカウント数の比率が所定のしきい値以下である場合には、その代表色を削除する処理をおこなう。削除された代表色に置換していた画素に対しては、残りの代表色からユークリッド距離の小さい代表色を新たに探索し、探索された代表色で置換する。
【００５８】
記憶部１６は、ハードディスク装置などの記憶装置であり、画像データ１６ａやパラメータデータ１６ｂなどを記憶する。ここで画像データ１６ａは、帳票などの画像データであり、パラメータデータ１６ｂは、さまざまな判定に用いられるしきい値などを記憶したデータである。制御部１７は、画像減色装置を全体制御する制御部であり、各機能部間の各種データの授受などを司る制御部である。
【００５９】
次に、図１に示した画像減色装置のハードウェア構成について説明する。図１２は、図１に示した画像減色装置のハードウェア構成を示す図である。同図に示すように、この画像減色装置は、キーボード１２０、ディスプレイ１２１、ＣＰＵ１２２、ＲＡＭ１２３、ＨＤＤ１２４およびＲＯＭ１２６をバス１２７で接続した構成となる。
【００６０】
ＨＤＤ１２４が格納および読み出し制御する記憶媒体であるハードディスク（ＨＤ）１２５には、本実施の形態で示される画像減色方法をコンピュータで実行することにより実現する画像減色プログラム１２５ａが記憶され、実行時にＲＡＭ１２３に読み込まれた後、ＣＰＵ１２２によりプログラムが解析され、画像減色プロセスの実行がおこなわれる。
【００６１】
この画像減色プロセスが、図１に示した色空間変換部１２、２次元ヒストグラム極大値探索部１３、１次元ヒストグラム極大値探索部１４および色置換部１５の各部の機能に対応する。また、画像データ１６ａやパラメータデータ１６ｂなどもＨＤ１２５に記憶され、ＲＡＭ１２３に読み込まれてＣＰＵ１２２により参照される。
【００６２】
次に、本実施の形態に係る画像減色装置がおこなう画像減色処理の処理手順について説明する。図１３は、本実施の形態に係る画像減色装置がおこなう画像減色処理の処理手順を示すフローチャートである。
【００６３】
同図に示すように、まず、式（１）から式（５）に基づいて、ＲＧＢ表色系で表されている画像データのＲＧＢ値をＹＣ表色系におけるＹＣ値に変換する（ステップＳ１３０１）。そして、ＹＣ表色系の色空間における３次元ヒストグラムから２次元ヒストグラムを作成する（ステップＳ１３０２）。
【００６４】
そして、作成された２次元ヒストグラムの平滑化をおこない（ステップＳ１３０３）、平滑化がおこなわれた２次元ヒストグラムの極大点を探索する（ステップＳ１３０４）。続いて、探索された２次元ヒストグラムの極大点を基にして、１次元ヒストグラムを作成し、作成された１次元ヒストグラムにおける極大点を探索する（ステップＳ１３０５）。この極大点は、画像に含まれる色を置換する代表色の候補となるものである。
【００６５】
その後、ステップＳ１３０３でおこなわれた平滑化処理が１回目の平滑化処理かどうかを調べ（ステップＳ１３０６）、１回目の平滑化処理であった場合には（ステップＳ１３０６，Ｙｅｓ）、ステップＳ１３０３に移行して、ステップＳ１３０３からステップＳ１３０５の処理を再度おこなう。
【００６６】
１回目の平滑化処理ではなかった場合には（ステップＳ１３０６，Ｎｏ）、前回平滑化処理をおこなった際に探索された極大点に対応する色と、今回平滑化処理をおこなった際に探索された極大点に対応する色との間のＹＣ色空間内でのユークリッド距離のうち、最大であるものを抽出し、その最大のユークリッド距離が所定のしきい値よりも大きいかどうかを調べる（ステップＳ１３０７）。所定のしきい値よりも大きくなかった場合には（ステップＳ１３０７，Ｎｏ）、ステップＳ１３０３に移行して、ステップＳ１３０３からステップＳ１３０６の処理を再度おこなう。
【００６７】
所定のしきい値よりも大きかった場合には（ステップＳ１３０７，Ｙｅｓ）、前回平滑化処理をおこなった際に探索された極大点に対応する色を代表色として選定し、それら代表色どうしの類似度を算出して、類似している場合に代表色を統合する処理をおこなう（ステップＳ１３０８）。その後、もとの画像に含まれる各画素の色を代表色で置換して（ステップＳ１３０９）、この画像減色処理を終了する。
【００６８】
次に、図１３に示した処理Ａの処理手順を詳細に説明する。図１４は、図１３に示した処理Ａの詳細な処理手順を示すフローチャートである。同図に示すように、処理Ａではまず、ＹＣ表色系に変換された画像の各画素の輝度Ｙの値に対して、水平方向のソーベルフィルタを適用し、式（７）を用いてｘｂｕｆの値を算出する（ステップＳ１４０１）。同様に、各画素の輝度Ｙの値に対して、垂直方向のソーベルフィルタを適用し、式（８）を用いてｙｂｕｆの値を算出する（ステップＳ１４０２）。
【００６９】
続いて、式（６）を用いて、エッジ強度ｗ１の値を算出し（ステップＳ１４０３）、そのエッジ強度ｗ１から、ヒストグラム加算値ｗを式（９）により算出する（ステップＳ１４０４）。そして、色空間内の座標点（Ｃ１，Ｃ２，Ｙ）に対応するすべての画素のヒストグラム加算値ｗを足し合わせて３次元ヒストグラムを作成するとともに、おなじ色差の座標（Ｃ１，Ｃ２）を有する点どうしでその３次元ヒストグラムの度数の値を累積した２次元ヒストグラムを作成する処理をおこなう（ステップＳ１４０５）。
【００７０】
その後、全画素に対してステップＳ１４０１からステップＳ１４０５までの処理が終了したかどうかを調べ（ステップＳ１４０６）、まだ処理が終了していない画素がある場合には（ステップＳ１４０６，Ｎｏ）、ステップＳ１４０１に移行して処理を実行する。すべての画素について処理が終了した場合には（ステップＳ１４０６，Ｙｅｓ）、そのまま処理Ａを終了する。
【００７１】
次に、図１３に示した処理Ｂの処理手順を詳細に説明する。図１５は、図１３に示した処理Ｂの詳細な処理手順を示すフローチャートである。同図に示すように、処理Ｂではまず、注目しているＣ１−Ｃ２平面上の色差の点（Ｃ１，Ｃ２）を取り囲む近傍８点における２次元ヒストグラムの度数の値を取得する（ステップＳ１５０１）。
【００７２】
そして、注目している色差の点の度数の値が、近傍８点の度数の値よりも大きいかどうかを調べ（ステップＳ１５０２）、近傍８点の度数の値よりも大きくない場合には（ステップＳ１５０２，Ｎｏ）、ステップＳ１５０５に移行する。近傍８点の度数の値よりも大きい場合には（ステップＳ１５０２，Ｙｅｓ）、注目している色差の点の度数の値が、Ｃ１−Ｃ２平面上の全点の度数の値を足し合わせた値の０．０００１％以上であるかどうかを調べる（ステップＳ１５０３）。
【００７３】
度数の値が０．０００１％以上ではない場合には（ステップＳ１５０３，Ｎｏ）、ステップＳ１５０５に移行する。度数の値が０．０００１％以上である場合には（ステップＳ１５０３，Ｙｅｓ）、その注目している色差の点（Ｃ１，Ｃ２）を、極大点として抽出する（ステップＳ１５０４）。
【００７４】
そして、すべての色差の点に対してステップＳ１５０１からステップＳ１５０４までの処理が終了したかどうかを調べ（ステップＳ１５０５）、まだ処理が終了していない色差の点がある場合には（ステップＳ１５０５，Ｎｏ）、ステップＳ１５０１に移行して処理を実行する。すべての色差の点について処理が終了した場合には（ステップＳ１５０５，Ｙｅｓ）、そのまま処理Ｂを終了する。
【００７５】
次に、図１３に示した処理Ｃの処理手順を詳細に説明する。図１６は、図１３に示した処理Ｃの詳細な処理手順を示すフローチャートである。同図に示すように、処理Ｃではまず、Ｃ１−Ｃ２平面に係る２次元ヒストグラムにおける極大点を通り、Ｃ１−Ｃ２平面に垂直な直線に対する１次元ヒストグラムを作成する（ステップＳ１６０１）。１次元ヒストグラムを作成する際には、式（１２）および式（１３）を用いて、直線上の各点に対応する度数の値を算出する。
【００７６】
そして、注目している直線上の点の度数の値が、両側の点の度数の値よりも大きいかどうかを調べ（ステップＳ１６０２）、両側の点の度数の値よりも大きくない場合には（ステップＳ１６０２，Ｎｏ）、ステップＳ１６０５に移行する。両側の点の度数の値よりも大きい場合には（ステップＳ１６０２，Ｙｅｓ）、注目している点の度数の値が、その直線上の全点の度数の値を足し合わせた値の０．００１％以上であるかどうかを調べる（ステップＳ１６０３）。
【００７７】
度数の値が０．００１％以上ではない場合には（ステップＳ１６０３，Ｎｏ）、ステップＳ１６０５に移行する。度数の値が０．００１％以上である場合には（ステップＳ１６０３，Ｙｅｓ）、その注目している点を、極大点として抽出する（ステップＳ１６０４）。
【００７８】
その後、直線上の全点に対してステップＳ１６０２からステップＳ１６０４で示した極大点の抽出処理が終了したかどうかを調べ（ステップＳ１６０５）、まだ処理が終了していない点がある場合には（ステップＳ１６０５，Ｎｏ）、ステップＳ１６０２に移行して処理を実行する。すべての点に対して処理が終了した場合には（ステップＳ１６０５，Ｙｅｓ）、２次元ヒストグラムの全極大点に対して、ステップＳ１６０１からステップＳ１６０５の処理が終了したかどうかを調べる（ステップＳ１６０６）。
【００７９】
全極大点に対する処理が終了していない場合には（ステップＳ１６０６，Ｎｏ）、ステップＳ１６０１に移行して処理を実行する。全極大点に対する処理が終了した場合には（ステップＳ１６０６，Ｙｅｓ）、そのまま処理Ｃを終了する。
【００８０】
次に、図１３に示した処理Ｄの処理手順を詳細に説明する。図１７は、図１３に示した処理Ｄの詳細な処理手順を示すフローチャートである。同図に示すように、処理Ｄではまず、１次元ヒストグラムの極大点に対応する色（代表色の候補色）のすべての組に対するユークリッド距離を算出する（ステップＳ１７０１）。
【００８１】
そして、各色の組に対する評価値を式（１４）により算出する（ステップＳ１７０２）。続いて、評価値が最小となる色の組を検索し（ステップＳ１７０３）、検索された評価値が、あらかじめ定められたしきい値よりも大きいかどうかを調べる（ステップＳ１７０４）。
【００８２】
評価値がしきい値よりも大きくない場合には（ステップＳ１７０４，Ｎｏ）、代表色の候補色の全度数に対する度数の比率が小さい方の色を除外して、比率が小さい方の色を大きい方の色に統合し（ステップＳ１７０５）、ステップＳ１７０３に移行する。評価値がしきい値よりも大きい場合には（ステップＳ１７０４，Ｙｅｓ）、そのまま処理Ｄを終了する。
【００８３】
次に、図１３に示した処理Ｅの処理手順を詳細に説明する。図１８は、図１３に示した処理Ｅの詳細な処理手順を示すフローチャートである。同図に示すように、処理Ｅではまず、画像の１画素おきに、抽出された極大点に対応する代表色と画素の色との間のユークリッド距離をＹＣ色空間において算出し、ユークリッド距離が最小となる代表色で画素の色を置換する（ステップＳ１８０１）。そして、代表色別に置換をおこなった画素数をカウントする（ステップＳ１８０２）。
【００８４】
続いて、１画素おきのすべての画素に対して、ステップＳ１８０１およびステップＳ１８０２の処理が終了したかどうかを調べ（ステップＳ１８０３）、すべての画素に対して処理が終了していない場合には（ステップＳ１８０３，Ｎｏ）、ステップＳ１８０１に移行して処理を実行する。
【００８５】
すべての画素に対して処理が終了した場合には（ステップＳ１８０３，Ｙｅｓ）、代表色別のカウント数が、すべての代表色のカウント数を足し合わせた数の０．１％よりも大きいかどうかを調べる（ステップＳ１８０４）。カウント数が０．１％よりも大きくない場合には（ステップＳ１８０４，Ｎｏ）、大きくなかった代表色を除外して（ステップＳ１８０６）、ステップＳ１８０５に移行する。これにより、色を置換する画素数があまりに少ない代表色を除外する。
【００８６】
カウント数が０．１％よりも大きい場合には（ステップＳ１８０４，Ｙｅｓ）、すべての代表色に対してステップＳ１８０４やステップＳ１８０６の処理が終了したかどうかを調べ（ステップＳ１８０５）、すべての代表色に対して処理が終了していない場合には（ステップＳ１８０５，Ｎｏ）、ステップＳ１８０４に移行して処理を実行する。
【００８７】
すべての代表色に対して処理が終了した場合には（ステップＳ１８０５，Ｙｅｓ）、注目している画素が、代表色に置換されていない画素、または、代表色が除外された代表色に置換されている画素かどうかを調べ（ステップＳ１８０７）、代表色に置換されていない画素、または、除外された代表色に置換されている画素でない場合には（ステップＳ１８０７，Ｎｏ）、ステップＳ１８０９に移行する。
【００８８】
代表色に置換されていない画素、または、除外された代表色に置換されている画素である場合には（ステップＳ１８０７，Ｙｅｓ）、画素の色を、その画素の色との間のＹＣ色空間におけるユークリッド距離が最小となる代表色で置換する（ステップＳ１８０８）。
【００８９】
その後、すべての画素に対してステップＳ１８０７やステップＳ１８０８の処理が終了したかどうかを調べ（ステップＳ１８０９）、まだ処理が終了していない画素がある場合には（ステップＳ１８０９，Ｎｏ）、ステップＳ１８０７に移行して処理を実行する。すべての画素について処理が終了した場合には（ステップＳ１８０９，Ｙｅｓ）、そのまま処理Ｅを終了する。
【００９０】
図１９は、本実施の形態に係る画像減色処理により処理された画像の一例を示す図である。図１９（ａ）は、ＲＧＢ表色系で表されたビットマップ画像であり、図１９（ｂ）は、図１９（ａ）のビットマップ画像から作成されたＪＰＥＧ（Ｊｏｉｎｔ Ｐｈｏｔｏｇｒａｐｈｉｃ Ｅｘｐｅｒｔｓ Ｇｒｏｕｐ）規格の画像であり、図１９（ｃ）は、図１９（ａ）のビットマップ画像から作成されたＪＰＥＧ２０００規格の画像であり、図１９（ｄ）は、本実施の形態で示した画像減色方法により図１９（ａ）のビットマップ画像から作成されたＰＮＧ（Ｐｏｒｔａｂｌｅ Ｎｅｔｗｏｒｋ Ｇｒａｐｈｉｃｓ）規格の画像である。
【００９１】
同図（ｄ）に示すように、本実施の形態に係る画像減色方法は、画像を作成する処理速度が高速なだけでなく、ＪＰＥＧ規格やＪＰＥＧ２０００規格の画像よりもくっきりと画像が表示されるため、特に、スキャナーなどでスキャンした帳票などの画像の処理を高速におこないたいような場合に適している。
【００９２】
以上説明してきたように、本実施の形態では、Ｃ１−Ｃ２平面に係る２次元ヒストグラムを色空間内の色の分布に基づいて作成し、作成された２次元ヒストグラムの極大値を探索し、探索された極大値に対応する点を通り、Ｃ１−Ｃ２平面に垂直な直線に係る１次元ヒストグラムを作成し、作成された１次元ヒストグラムの極大値を探索し、探索された１次元ヒストグラムの極大値に対応する色を代表色として選定し、選定された代表色で画像に含まれる色を置換することとしたので、ＹＣ表色系の色空間内で３次元的に代表色の探索をおこなうのではなく、２次元探索と１次元探索に分割して探索をおこなうことにより、短い処理時間で効率的に代表色の選定をおこなうことができる。
【００９３】
また、Ｃ１−Ｃ２平面に係る２次元ヒストグラムに対して平滑化処理を適切におこなうこととしたので、代表色の選定を的確におこなうことができる。
【００９４】
さて、これまで本発明の実施の形態について説明したが、本発明は上述した実施の形態以外にも、上記特許請求の範囲に記載した技術的思想の範囲内において種々の異なる実施の形態にて実施されてもよいものである。
【００９５】
たとえば、本実施の形態では、ＲＧＢ表色系の画像をＹＣ表色系の画像に変換し、変換された画像に対して減色処理をおこなうこととしたが、本発明はこれに限定されるものではなく、明度Ｌ、緑から赤の範囲の色度の要素ａ、および、青から黄の範囲の色度の要素ｂからなるＬａｂ表色系の画像などに対しても同様の処理を適用することができる。
【００９６】
また、本実施の形態では、２次元ヒストグラムの平滑化をおこなって、平滑化された２次元ヒストグラムの極大値を探索することとしたが、本発明はこれに限定されるものではなく、平滑化をおこなわずに極大点を探索し、探索された極大点に対応する色を統合するなどして代表色として選定することとしてもよい。このように、平滑化をおこなわないことにより、より短い時間で代表色の選定をおこなうことができる。
【００９７】
また、本実施の形態において説明した各処理のうち、自動的におこなわれるものとして説明した処理の全部または一部を手動的におこなうこともでき、あるいは、手動的におこなわれるものとして説明した処理の全部または一部を公知の方法で自動的におこなうこともできる。この他、上記文書中や図面中で示した処理手順、制御手順、具体的名称、各種のデータやパラメータを含む情報については、特記する場合を除いて任意に変更することができる。
【００９８】
また、図示した各装置の各構成要素は機能概念的なものであり、必ずしも物理的に図示のように構成されていることを要しない。すなわち、各装置の分散・統合の具体的形態は図示のものに限られず、その全部または一部を、各種の負荷や使用状況などに応じて、任意の単位で機能的または物理的に分散・統合して構成することができる。さらに、各装置にて行なわれる各処理機能は、その全部または任意の一部が、ＣＰＵおよび当該ＣＰＵにて解析実行されるプログラムにて実現され、あるいは、ワイヤードロジックによるハードウェアとして実現され得る。
【００９９】
なお、本実施の形態で説明した画像減色方法は、図１２において説明したように、あらかじめ用意されたプログラムをパーソナル・コンピュータやワークステーションなどのコンピュータで実行することによって実現することができる。このプログラムは、インターネットなどのネットワークを介して配布することができる。また、このプログラムは、ハードディスク、フレキシブルディスク（ＦＤ）、ＣＤ−ＲＯＭ、ＭＯ、ＤＶＤなどのコンピュータで読み取り可能な記録媒体に記録され、コンピュータによって記録媒体から読み出されることによって実行することもできる。
【０１００】
【発明の効果】
以上説明したように、請求項１の発明によれば、画像に含まれる色の特性を３次元の座標に置き換えた色空間の所定の平面に係る２次元ヒストグラムを色空間内における画像に含まれる色の分布に基づいて作成して、作成された２次元ヒストグラムの極大値を探索し、探索された極大値に対応する点を通り、色空間の所定の平面に垂直な直線に係る１次元ヒストグラムを色空間内における画像に含まれる色の分布に基づいて作成して、作成された１次元ヒストグラムの極大値を探索し、探索された１次元ヒストグラムの極大値に対応する色を代表色として選定し、選定された代表色で画像に含まれる色を置換することとしたので、色空間内で３次元的に代表色の探索をおこなうのではなく、２次元探索と１次元探索に分割して探索をおこなって代表色を選定することにより、短い処理時間で効率的にデータサイズの小さい画像を作成することができる。
【０１０１】
また、請求項２の発明によれば、画像のエッジ強度を考慮した２次元ヒストグラムを色空間内における画像に含まれる色の分布に基づいて作成し、作成された２次元ヒストグラムの極大値を探索することとしたので、エッジ部分に生じるノイズの色を代表色として選定することを効率的に抑制することができる。
【０１０２】
また、請求項３の発明によれば、極大値に対応する点を通り、色空間の所定の平面に垂直な直線上の点から色空間の所定の距離内にある点に対応する色の分布に基づいて、所定の平面に垂直な直線に係る１次元ヒストグラムを作成し、作成された１次元ヒストグラムの極大値を探索することとしたので、色空間内の色の分布を調べることにより、より密な分布を示す代表色を効率的に抽出することができる。
【０１０３】
また、請求項４の発明によれば、色空間の所定の平面に係る２次元ヒストグラムを色空間内における画像に含まれる色の分布に基づいて作成し、作成された２次元ヒストグラムの平滑化をおこなって、平滑化された２次元ヒストグラムの極大値を探索し、探索された極大値に対応する点を通り、所定の平面に垂直な直線に係る１次元ヒストグラムを色空間内における画像に含まれる色の分布に基づいて作成し、作成された１次元ヒストグラムの極大値を探索し、平滑化前および平滑化後の１次元ヒストグラムの極大値に対応する色間の類似度を算出して、算出された類似度に基づいて２次元ヒストグラムの平滑化、平滑化された前記２次元ヒストグラムの極大値の探索、および、１次元ヒストグラムの極大値の探索とをさらにおこなうかどうかを判定することとしたので、平滑化をおこない、かつ、平滑化をおこなう回数を平滑化前後の色間の類似度に基づいて判定することにより、抽出される極大点の数を制御して、適切な数の代表色を得ることができる。
【０１０４】
また、請求項５の発明によれば、探索された１次元ヒストグラムの複数の極大値に対応する色を色の類似度に基づいて統合し、統合された色を代表色として選定し、選定された代表色で画像に含まれる色を置換することとしたので、類似した色を統合することにより、さらにデータサイズの小さい画像を効率的に作成することができる。
【０１０５】
また、請求項６の発明によれば、画像に含まれる色の特性を３次元の座標に置き換えた色空間の所定の平面に係る２次元ヒストグラムを色空間内における画像に含まれる色の分布に基づいて作成して、作成された２次元ヒストグラムの極大値を探索し、探索された極大値に対応する点を通り、色空間の所定の平面に垂直な直線に係る１次元ヒストグラムを色空間内における画像に含まれる色の分布に基づいて作成して、作成された１次元ヒストグラムの極大値を探索し、探索された１次元ヒストグラムの極大値に対応する色を代表色として選定し、選定された代表色で画像に含まれる色を置換することとしたので、色空間内で３次元的に代表色の探索をおこなうのではなく、２次元探索と１次元探索に分割して探索をおこなって代表色を選定することにより、短い処理時間で効率的にデータサイズの小さい画像を作成することができる。
【０１０６】
また、請求項７の発明によれば、画像のエッジ強度を考慮した２次元ヒストグラムを色空間内における画像に含まれる色の分布に基づいて作成し、作成された２次元ヒストグラムの極大値を探索することとしたので、エッジ部分に生じるノイズの色を代表色として選定することを効率的に抑制することができる。
【０１０７】
また、請求項８の発明によれば、極大値に対応する点を通り、色空間の所定の平面に垂直な直線上の点から色空間の所定の距離内にある点に対応する色の分布に基づいて、所定の平面に垂直な直線に係る１次元ヒストグラムを作成し、作成された１次元ヒストグラムの極大値を探索することとしたので、色空間内の色の分布を調べることにより、より密な分布を示す代表色を効率的に抽出することができる。
【０１０８】
また、請求項９の発明によれば、色空間の所定の平面に係る２次元ヒストグラムを色空間内における画像に含まれる色の分布に基づいて作成し、作成された２次元ヒストグラムの平滑化をおこなって、平滑化された２次元ヒストグラムの極大値を探索し、探索された極大値に対応する点を通り、所定の平面に垂直な直線に係る１次元ヒストグラムを色空間内における画像に含まれる色の分布に基づいて作成し、作成された１次元ヒストグラムの極大値を探索し、平滑化前および平滑化後の１次元ヒストグラムの極大値に対応する色間の類似度を算出して、算出された類似度に基づいて２次元ヒストグラムの平滑化、平滑化された前記２次元ヒストグラムの極大値の探索、および、１次元ヒストグラムの極大値の探索とをさらにおこなうかどうかを判定することとしたので、平滑化をおこない、かつ、平滑化をおこなう回数を平滑化前後の色間の類似度に基づいて判定することにより、抽出される極大点の数を制御して、適切な数の代表色を得ることができる。
【０１０９】
また、請求項１０の発明によれば、探索された１次元ヒストグラムの複数の極大値に対応する色を色の類似度に基づいて統合し、統合された色を代表色として選定し、選定された代表色で画像に含まれる色を置換することとしたので、類似した色を統合することにより、さらにデータサイズの小さい画像を効率的に作成することができる。
【０１１０】
また、請求項１１の発明によれば、画像に含まれる色の特性を３次元の座標に置き換えた色空間の所定の平面に係る２次元ヒストグラムを色空間内における画像に含まれる色の分布に基づいて作成して、作成された２次元ヒストグラムの極大値を探索し、探索された極大値に対応する点を通り、色空間の所定の平面に垂直な直線に係る１次元ヒストグラムを色空間内における画像に含まれる色の分布に基づいて作成して、作成された１次元ヒストグラムの極大値を探索し、探索された１次元ヒストグラムの極大値に対応する色を代表色として選定し、選定された代表色で画像に含まれる色を置換することとしたので、色空間内で３次元的に代表色の探索をおこなうのではなく、２次元探索と１次元探索に分割して探索をおこなって代表色を選定することにより、短い処理時間で効率的にデータサイズの小さい画像を作成することができる。
【図面の簡単な説明】
【図１】本実施の形態に係る画像減色装置の構成を示す機能ブロック図である。
【図２】本実施の形態に係る画像減色装置により減色処理がなされるＲＧＢ画像の一例を示す図である。
【図３】ＲＧＢ表色系からＹＣ表色系に変換して得られた画像データの色空間の一例を示す図である。
【図４】各画素の重み付けに用いるソーベルフィルタのオペレータを示す図である。
【図５】Ｃ１−Ｃ２平面における２次元ヒストグラムの一例を示す図である。
【図６】Ｃ１−Ｃ２平面に係る２次元ヒストグラムに適用される平滑化フィルタのオペレータを示す図である。
【図７】Ｃ１−Ｃ２平面に係る２次元ヒストグラムの極大点を通るＣ１−Ｃ２平面に垂直な直線の一例を示す図である。
【図８】１次元ヒストグラムの作成方法を説明する概念図である。
【図９】１次元ヒストグラムにおける極大点の概念を示す図である。
【図１０】平滑化処理による極大点の数および代表色候補の色数の変化を示す図である。
【図１１】平滑化処理がなされたＣ１−Ｃ２平面における２次元ヒストグラムの一例を示す図である。
【図１２】図１に示した画像減色装置のハードウェア構成を示す図である。
【図１３】本実施の形態に係る画像減色装置がおこなう画像減色処理の処理手順を示すフローチャートである。
【図１４】図１３に示した処理Ａの詳細な処理手順を示すフローチャートである。
【図１５】図１３に示した処理Ｂの詳細な処理手順を示すフローチャートである。
【図１６】図１３に示した処理Ｃの詳細な処理手順を示すフローチャートである。
【図１７】図１３に示した処理Ｄの詳細な処理手順を示すフローチャートである。
【図１８】図１３に示した処理Ｅの詳細な処理手順を示すフローチャートである。
【図１９】本実施の形態に係る画像減色処理により処理された画像の一例を示す図である。
【符号の説明】
１０ 入力部
１１ 表示部
１２ 色空間変換部
１３ ２次元ヒストグラム極大値探索部
１４ １次元ヒストグラム極大値探索部
１５ 色置換部
１６ 記憶部
１６ａ，１２３ｂ 画像データ
１６ｂ，１２３ｃ パラメータデータ
１２０ キーボード
１２１ ディスプレイ
１２２ ＣＰＵ
１２３ ＲＡＭ
１２３ａ，１２５ａ 画像減色プログラム
１２４ ＨＤＤ
１２５ ＨＤ
１２６ ＲＯＭ
１２７ バス[0001]
BACKGROUND OF THE INVENTION
The present invention relates to an image color reduction device that selects a representative color that represents a color included in an image and replaces the color included in the image with a representative color to create an image with a reduced number of colors in the image, and an image color reduction More particularly, the present invention relates to an image color reduction apparatus, an image color reduction method, and an image color reduction program capable of efficiently creating an image with a small data size in a short processing time.
[0002]
[Prior art]
Conventionally, when a document or form is scanned with a scanner and the image is digitized so that it can be handled by a computer, the color of the image can be expressed by a combination of the three primary colors red, green, and blue (RGB). It is common. For example, in a digital image in which each color of red, green, and blue is divided into 256 stages and gradation information of each color is held in 8 bits, tone expression of approximately 16.77 million colors can be performed.
[0003]
In desktop publishing and image processing systems, processing such as processing and correction of digital images can be performed. However, since the amount of image information is very large, processing takes time, and image data The size will increase.
[0004]
Therefore, Patent Document 1 discloses a method of detecting a similar color area in an image and replacing a color in the detected area with one dominant color in the area. According to this method, a dominant color that creates a three-dimensional histogram representing the distribution of colors in an image with the gradation values of red, green, and blue as axes, and performs replacement based on the three-dimensional histogram. To decide.
[0005]
Then, after the dominant color is determined, the original image is scanned along the scanning line to investigate the adjacency relationship between a plurality of regions having similar colors, and the regions having similar colors are adjacent to each other. If they match, the regions are combined, and the color of the combined region is replaced with the dominant color determined based on the three-dimensional histogram.
[0006]
[Patent Document 1]
Japanese Patent No. 3299778
[0007]
[Problems to be solved by the invention]
However, the conventional technique of Patent Document 1 has a problem in that it takes time to detect a similar color region and replace the color of the region with a dominant color, which is not efficient.
[0008]
Specifically, when detecting regions of similar colors in the image, in order to determine the dominant color by searching the color distribution three-dimensionally in the positive and negative directions of each axis of the three-dimensional histogram, There was a problem that processing took time.
[0009]
In addition, after deciding the dominant color, we investigate the adjacency relationship between multiple areas with similar colors and combine them when they are adjacent, but the color change is severe, In the case of a complex image, there has been a problem that it takes a long time to perform a process of investigating adjacent relations and combining them.
[0010]
The present invention has been made to solve the above problems (problems). An image color reduction apparatus, an image color reduction method, and an image color reduction program capable of efficiently creating an image with a small data size in a short processing time. The purpose is to provide.
[0011]
[Means for Solving the Problems]
In order to achieve the above object, an image color reduction device according to the invention of claim 1 selects a representative color representing a color included in an image, replaces the color included in the image with the representative color, and An image color reduction apparatus for creating an image with a reduced number of colors of an image, wherein a two-dimensional histogram relating to a predetermined plane of a color space obtained by replacing color characteristics included in the image with three-dimensional coordinates is displayed in the color space. A two-dimensional histogram maximum value search means for searching for a maximum value of the generated two-dimensional histogram, and a maximum value searched by the two-dimensional histogram maximum value search means A one-dimensional histogram that passes through a point corresponding to the value and is related to a straight line perpendicular to a predetermined plane of the color space based on a distribution of colors included in the image in the color space; The poles of One-dimensional histogram maximum value searching means for searching for a value, and a color corresponding to the maximum value of the one-dimensional histogram searched by the one-dimensional histogram maximum value searching means is selected as the representative color, and the selected representative color And color replacement means for replacing colors included in the image.
[0012]
In the image color reduction device according to the second aspect of the present invention, the two-dimensional histogram maximum value search means converts the two-dimensional histogram considering the edge strength of the image into a distribution of colors included in the image in the color space. And generating a maximum value of the created two-dimensional histogram.
[0013]
Further, in the image color reduction device according to the invention of claim 3, the one-dimensional histogram maximum value searching means passes through a point corresponding to the maximum value searched by the two-dimensional histogram maximum value searching means, and the predetermined color of the color space is determined. A one-dimensional histogram relating to a straight line perpendicular to the predetermined plane is created based on a color distribution corresponding to a point within a predetermined distance of the color space from a point on the straight line perpendicular to the plane. Further, it is characterized by searching for the maximum value of the one-dimensional histogram.
[0014]
According to a fourth aspect of the present invention, there is provided the image color reduction apparatus, wherein the two-dimensional histogram maximum value search means uses a two-dimensional histogram relating to a predetermined plane of the color space as a distribution of colors included in the image in the color space. The two-dimensional histogram is smoothed and searched for the maximum value of the smoothed two-dimensional histogram, and the one-dimensional histogram maximum value search means is configured to search for the two-dimensional histogram maximum value. A one-dimensional histogram relating to a straight line that passes through a point corresponding to the maximum value searched by the search means and is perpendicular to the predetermined plane is generated based on the distribution of colors included in the image in the color space. And searching for a maximum value of the one-dimensional histogram, calculating a similarity between colors corresponding to the maximum value of the one-dimensional histogram before and after the smoothing, and calculating the similarity Based on the smoothing of the two-dimensional histogram by the two-dimensional histogram maximum value search means, the search for the maximum value of the smoothed two-dimensional histogram, and the search for the maximum value of the one-dimensional histogram by the one-dimensional histogram maximum value search means It is characterized by determining whether or not to perform further.
[0015]
Further, in the image color reduction device according to the invention of claim 5, the color replacement means assigns colors corresponding to a plurality of maximum values of the one-dimensional histogram searched by the one-dimensional histogram maximum value searching means to the similarity of the colors. And the integrated color is selected as the representative color, and the color included in the image is replaced with the selected representative color.
[0016]
The image color reduction method according to the invention of claim 6 reduces the number of colors of the image by selecting a representative color representative of the color included in the image and replacing the color included in the image with the representative color. An image color reduction method for creating a processed image, wherein the image in the color space includes a two-dimensional histogram relating to a predetermined plane of the color space in which the color characteristics included in the image are replaced with three-dimensional coordinates. A two-dimensional histogram maximum value searching step for searching for a maximum value of the generated two-dimensional histogram, and a point corresponding to the maximum value searched by the two-dimensional histogram maximum value searching step. As described above, a one-dimensional histogram relating to a straight line perpendicular to a predetermined plane of the color space is created based on the distribution of colors included in the image in the color space, and a maximum value of the created one-dimensional histogram is searched. One dimension A color corresponding to the maximum value of the one-dimensional histogram searched in the strogram maximum value search step and the one-dimensional histogram maximum value search step is selected as the representative color, and the color included in the image with the selected representative color And a color replacement step for replacing.
[0017]
In the image color reduction method according to the invention of claim 7, in the two-dimensional histogram maximum value search step, the two-dimensional histogram in consideration of the edge strength of the image is converted into a distribution of colors included in the image in the color space. And generating a maximum value of the created two-dimensional histogram.
[0018]
In the image color reduction method according to an eighth aspect of the present invention, the one-dimensional histogram maximum value search step passes through a point corresponding to the maximum value searched by the two-dimensional histogram maximum value search step, and the predetermined color of the color space is determined. A one-dimensional histogram relating to a straight line perpendicular to the predetermined plane is created based on a color distribution corresponding to a point within a predetermined distance of the color space from a point on the straight line perpendicular to the plane. Further, it is characterized by searching for the maximum value of the one-dimensional histogram.
[0019]
Further, in the image color reduction method according to the invention of claim 9, in the two-dimensional histogram maximum value searching step, a distribution of colors included in the image in the color space is obtained from a two-dimensional histogram related to a predetermined plane of the color space. The two-dimensional histogram thus created is smoothed to find a maximum value of the smoothed two-dimensional histogram, and the one-dimensional histogram maximum value search step includes the two-dimensional histogram maximum value. A one-dimensional histogram relating to a straight line that passes through a point corresponding to the maximum value searched in the search step and is perpendicular to the predetermined plane is created based on the distribution of colors included in the image in the color space. And searching for a maximum value of the one-dimensional histogram, calculating a similarity between colors corresponding to the maximum value of the one-dimensional histogram before and after the smoothing, and calculating the similarity Based on the smoothing of the two-dimensional histogram by the two-dimensional histogram maximum value search step, the search of the maximum value of the smoothed two-dimensional histogram, and the search of the maximum value of the one-dimensional histogram by the one-dimensional histogram maximum value search step It is characterized by determining whether or not to perform further.
[0020]
Further, in the image color reduction method according to the invention of claim 10, in the color replacement step, colors corresponding to a plurality of maximum values of the one-dimensional histogram searched in the one-dimensional histogram maximum value search step are calculated. And the integrated color is selected as the representative color, and the color included in the image is replaced with the selected representative color.
[0021]
The image color reduction program according to the invention of claim 11 reduces the number of colors of the image by selecting a representative color representative of the color included in the image and replacing the color included in the image with the representative color. An image color reduction program for creating a processed image, wherein the image in the color space includes a two-dimensional histogram relating to a predetermined plane of the color space in which the color characteristics included in the image are replaced with three-dimensional coordinates A two-dimensional histogram maximum value search procedure for searching for a maximum value of the generated two-dimensional histogram, and a point corresponding to the maximum value searched by the two-dimensional histogram maximum value search procedure. As described above, a one-dimensional histogram relating to a straight line perpendicular to a predetermined plane of the color space is created based on the distribution of colors included in the image in the color space, and the maximum value of the created one-dimensional histogram is obtained. One-dimensional histogram maximum value searching procedure to be searched and a color corresponding to the maximum value of the one-dimensional histogram searched by the one-dimensional histogram maximum value searching procedure is selected as the representative color, and the image is selected with the selected representative color. And a color replacement procedure for replacing colors included in the computer.
[0022]
DETAILED DESCRIPTION OF THE INVENTION
Exemplary embodiments of an image color reduction apparatus, an image color reduction method, and an image color reduction program according to the present invention will be explained below in detail with reference to the accompanying drawings. In the present embodiment, image data expressed in the RGB color system is converted into a YC color system that expresses a color with a luminance component and two color difference components, and image color reduction processing is performed in the YC color system. Let's show the case to do. Since the human optic nerve is more sensitive to changes in luminance than changes in color, the YC color system is a color system that takes into account such characteristics of the optic nerve.
[0023]
First, the functional configuration of the image color reduction device according to the present embodiment will be described. FIG. 1 is a functional block diagram showing a configuration of an image color reduction apparatus according to the present embodiment. This image color reduction device replaces the color of an image acquired from a scanner or the like with a representative color that represents the color of the image and performs color reduction processing in a three-dimensional representative color in the color space of the YC color system. Instead of performing a search, a representative color is efficiently selected in a short processing time by dividing into a two-dimensional search and a one-dimensional search. In addition, even if an image has a drastic color change, it is replaced with a representative color for each pixel, so there is no problem in processing time, and the selection of the representative color can be accurately performed by appropriately smoothing the image. It is configured so that it can be performed.
[0024]
FIG. 2 is a diagram illustrating an example of an RGB image subjected to color reduction processing by the image color reduction device according to the present embodiment. This image color reduction apparatus is particularly suitable for color reduction processing of a color image such as a form digitized by a scanner. As shown in the figure, in this form image example, the background is white, the color of the item column such as customer number is yellow, the color of the frame of the entry column for entering numbers etc. is red, and The printed characters are black.
[0025]
However, each color is not uniform and there is usually a color blur. This relies on the scanner containing various noises when digitizing. For example, when a form having an edge portion whose color changes from white to red across a boundary line is digitized by a scanner, pink pixels are generated at the boundary portion. Therefore, in the color distribution of the digitized image, a distribution that continuously changes from white to red is observed.
[0026]
Returning to the description of FIG. 1, the image color reduction device includes an input unit 10, a display unit 11, a color space conversion unit 12, a two-dimensional histogram maximum value search unit 13, a one-dimensional histogram maximum value search unit 14, and a color replacement unit 15. A storage unit 16 and a control unit 17. The input unit 10 is an input device such as a keyboard or a mouse, and the display unit 11 is a display device such as a display.
[0027]
The color space conversion unit 12 is a conversion unit that converts the color space of image data represented in the RGB color system to the YC color system. In particular,

Is converted from the RGB color system to the YC color system. Here, Y is a luminance value in the YC color system, C1 ′ and C2 ′ are values of two color differences in the YC color system, and R, G, and B are RGB color system values, respectively. Represents the gradation values of red, green and blue.
[0028]
Further, the color space conversion unit 12 can express the color differences C1 ′ and C2 ′ with a data amount of 1 byte.
C1 = max (0, min (255, C1 ′ + 128)). . . . . . (4)
C2 = max (0, min (255, C2 ′ + 128)). . . . . . (5)
The values of the color differences C1 ′ and C2 ′ are converted using the relational expression. Here, max (a, b) means the larger value of a and b, and min (a, b) means the smaller value of a and b. The converted color differences C1 and C2 take values from 0 to 255.
[0029]
FIG. 3 shows an example of the color space of image data obtained by converting from the RGB color system to the YC color system. Color differences C1 and C2 are set on the horizontal axis and luminance Y is set on the vertical axis, and the color distribution is represented as a collection of points. Although not shown in FIG. 3, each point in the color space holds not only the coordinate value but also information on the frequency of the color distribution.
[0030]
Returning to the description of FIG. 1, the two-dimensional histogram maximum value search unit 13 creates a two-dimensional histogram related to the C1-C2 plane of the color space based on the color distribution converted by the color space conversion unit 12, It is a search part which searches for the maximum value of the created two-dimensional histogram.
[0031]
Specifically, first, in order to prevent the color of the edge portion of the image from being selected as the representative color, processing for creating a histogram in consideration of edge strength is performed. This is because, for example, when a form with a red area in a white area is scanned, the color of the edge of the red area of the image obtained by scanning is pink that is not in the original form. This is to suppress the selection of pink instead of red as a representative color in such a case.
[0032]
This edge strength is calculated by applying a Sobel filter to the value of the luminance Y of each pixel of the image. FIG. 4 is a diagram illustrating an operator of a Sobel filter used for weighting each pixel. 4A is an operator for detecting a vertical edge, and FIG. 4B is an operator for detecting a horizontal edge. FIG. 4C shows the gradation value of the luminance Y of the pixel of interest and its neighboring 8 pixels.
[0033]
By applying a Sobel filter, the edge strength w1 is

Can be calculated. Here, || means to take the absolute value of the value. Since the edge strength w1 is the sum of absolute values of values obtained by applying an operator for detecting edges in the vertical direction and the horizontal direction, edges in both the vertical and horizontal directions can be considered.
[0034]
Then, based on the calculated edge strength w1, the histogram addition value w of each pixel of the image is
w = max (0,1-w1 / edge). . . . . . (9)
Calculated by Here, max (a, b) means the larger value of a and b, and edge is a predetermined constant of 1 or more.
[0035]
Subsequently, the histogram addition values w of all the pixels corresponding to the coordinate point (C1, C2, Y) in the color space are added, and the added value is used as the frequency for the coordinate point (C1, C2, Y). Thus, a three-dimensional histogram h (C1, C2, Y) is created.
[0036]
Then, a two-dimensional histogram H relating to the C1-C2 plane of the color space ₁ (C1, C2)
H ₁ (C1, C2) = Σh (C1, C2, Y). . . . . . (10)
Create by. Here, Σ means taking the sum of the frequencies h (C1, C2, Y) of all points corresponding to the color difference coordinates designated by C1 and C2.
[0037]
Here, in Expression (9), the edge strength w1 is more strongly reflected in the histogram addition value w as the edge value is smaller. Conversely, as the edge value increases, the edge strength w1 value is not reflected in the histogram addition value w. Therefore, by appropriately setting the value of edge, it is possible to adjust the degree of not selecting the color of the edge portion of the image as the representative color.
[0038]
Two-dimensional histogram H related to C1-C2 plane ₁ After (C1, C2) is created, the two-dimensional histogram maximum value search unit 13 performs a two-dimensional histogram H ₁ The value of (C1, C2) is examined, and the frequency of the color difference point (C1, C2) being noticed is greater than the frequency of the eight neighboring points surrounding the color difference point on the C1-C2 plane, and When the frequency of the color difference point is 0.0001% or more of the total frequency (the sum of the frequencies of all points), the color difference point indicating the frequency is set as the maximum point indicating the maximum value of the frequency. Perform the extraction process. In this way, the threshold value for the frequency is set, and only those having a certain degree of frequency are selected as local maximum points.
[0039]
FIG. 5 is a diagram illustrating an example of a two-dimensional histogram in the C1-C2 plane. As shown in the figure, this two-dimensional histogram shows the values H of the frequency difference C1 and C2 taking the edge strength w1 into consideration on the horizontal axis. ₁ (C1, C2) is set on the vertical axis. Further, FIG. 5 shows the positions of eleven local maximum points indicated by US marks.
[0040]
Returning to the description of FIG. 1, the two-dimensional histogram maximum value search unit 13 further performs a smoothing process on the obtained two-dimensional histogram related to the C1-C2 plane. This is to control the number of maximum points extracted based on the two-dimensional histogram.
[0041]
This smoothing process is performed by applying a smoothing filter to the two-dimensional histogram relating to the C1-C2 plane. FIG. 6 is a diagram illustrating an operator of a smoothing filter applied to a two-dimensional histogram related to the C1-C2 plane. FIG. 6A is an operator applied to the color difference points (C1, C2) of interest and the color difference points of 8 points in the vicinity thereof, and FIG. 6B is a point of color differences of interest. The frequency values corresponding to the color difference points of (C1, C2) and its neighboring eight points are represented.
[0042]
By applying a smoothing filter, a two-dimensional histogram H corresponding to the color difference point (C1, C2) ₂ (C1, C2) is

Is calculated by After performing the smoothing process, the two-dimensional histogram maximum value search unit 13 performs the two-dimensional histogram H ₂ Search for the maximum point of (C1, C2).
[0043]
Returning to the description of FIG. 1, the one-dimensional histogram maximum value search unit 14 passes through the maximum point searched by the two-dimensional histogram maximum value search unit 13 and is perpendicular to the C1-C2 plane of the color space of the YC color system. A search unit that generates a one-dimensional histogram related to a straight line and searches for a maximum value of the generated one-dimensional histogram.
[0044]
FIG. 7 is a diagram illustrating an example of a straight line that passes through the maximum point of the two-dimensional histogram related to the C1-C2 plane and is perpendicular to the C1-C2 plane. As shown in the figure, these straight lines pass through the maximum points searched by the two-dimensional histogram maximum value search unit 13 and extend perpendicular to the C1-C2 plane.
[0045]
FIG. 8 is a conceptual diagram illustrating a method for creating a one-dimensional histogram. As shown in the figure, when creating a one-dimensional histogram, a straight line that passes through the maximum point in the color space and is perpendicular to the C1-C2 plane is considered, and each point K (C1, C2, k on the straight line is considered. ) H of the frequencies of points within a predetermined distance D from _Three (C1, C2, k)
H _Three (C1, C2, k) = Σ {h (C1, C2, k) × w2}. . . (12)
This value is used as the frequency of the one-dimensional histogram at the point K. Here, D ≦ k ≦ 255−D.
[0046]
H (C1, C2, k) is a three-dimensional histogram created by the histogram addition value of Equation (9), and w2 is a weight given to h (C1, C2, k). . This weight w2 is
w2 = exp (−L ² / 50-M ² / 2-N ² / 2). . . . . . (13)
Is calculated by Here, L, M, and N are distances in the Y-axis direction, the C1-axis direction, and the C2-axis direction, respectively, between the point within the predetermined distance D and the point K (C1, C2, k). It is. The weight w2 decreases as the distance from the point K (C1, C2, k) increases. Σ means taking the sum of values obtained by multiplying the frequency h (C1, C2, k) of all points within the predetermined distance D by the weight w2.
[0047]
FIG. 9 is a diagram showing the concept of the maximum point in the one-dimensional histogram. As shown in the figure, in this example, as a result of creating a one-dimensional histogram by the method described above, it can be seen that the two maximum points are on a straight line. By extracting this maximum point, it is possible to extract a color that is a candidate for a representative color that represents the color of the image.
[0048]
Specifically, in the one-dimensional histogram on the line segment corresponding to each local maximum point, the frequency is greater than the frequency of the adjacent points on the line segment, and the frequency is the frequency of each point on the line segment. When it is larger than 0.001% of the total, the point is extracted as a maximum point.
[0049]
The two-dimensional histogram smoothing process, the search process for the maximum point in the smoothed two-dimensional histogram, and the search process for the maximum point in the one-dimensional histogram satisfy a predetermined condition by the one-dimensional histogram search unit 14. It is repeated until it is determined that
[0050]
Specifically, the one-dimensional histogram search unit 14 stores colors corresponding to the local maximum points searched in the previous iteration, and colors corresponding to the newly searched local maximum points. The Euclidean distance in the color space between and is calculated, and a process for checking whether or not the maximum value of the Euclidean distance is larger than a preset threshold value is performed. When the maximum value of the Euclidean distance becomes larger than the threshold value, the local maximum point searched in the previous iteration is adopted as a representative color candidate, and the iteration process is terminated.
[0051]
Also, when the number of newly searched maximum points is less than a preset minimum number of maximum points, or when the number of repetitions exceeds a predetermined number, the previous one is repeated. The searched local maximum point is adopted as a representative color candidate, and the iterative process is terminated. In other cases, it is determined that integration of local maximum points is still possible, and the process is repeated.
[0052]
FIG. 10 is a diagram illustrating changes in the number of maximum points and the number of colors of representative color candidates due to the smoothing process. As shown in the figure, in this example, in the first smoothing, there are 11 local maximum points in the two-dimensional histogram, and the color corresponding to the local maximum point in the one-dimensional histogram, that is, a representative color candidate. There are 17 colors. However, as the number of times of smoothing increases, local maximum points having similar color difference characteristics are integrated, and the number of local maximum points and the number of colors that are candidates for representative colors decrease.
[0053]
FIG. 11 is a diagram illustrating an example of a two-dimensional histogram in the C1-C2 plane that has been subjected to the smoothing process. This figure shows a state in which the smoothing process is performed on the two-dimensional histogram having 11 local maximum points shown in FIG. 5 and the number of local maximum points is reduced to 5.
[0054]
Returning to the description of FIG. 1, the color replacement unit 15 selects a color corresponding to the maximum value of the one-dimensional histogram searched by the one-dimensional histogram maximum value search unit 14 as a representative color, and an image is displayed using the selected representative color. This is a replacement unit that replaces the colors included in.
[0055]
Specifically, the evaluation value E is assigned to all the sets of colors (representative candidate colors) corresponding to the searched local maximum points.
E = (Euclidean distance between candidate color i and candidate color j)
+ Α × min (the ratio of the frequency of candidate color i to the sum of the frequencies of all candidate colors,
The ratio of the frequency of candidate color j to the sum of the frequencies of all candidate colors)
. . . . . . (14)
Calculated by Here, (Euclidean distance between candidate color i and candidate color j) is the Euclidean distance between two colors in the color space (C1, C2, Y), and the frequency value of the candidate color is the frequency in the one-dimensional histogram. Min (a, b) means the smaller value of a and b, and α is a predetermined positive constant.
[0056]
When the evaluation value E is smaller than a predetermined threshold value, candidate colors having a small frequency ratio with respect to the sum of the frequencies of all candidate colors are deleted and integrated into candidate colors having a high frequency ratio. . Here, in the formula (14) for calculating the evaluation value E, the frequency ratio of candidate colors is considered, but this is because candidate colors having a large frequency value are deleted despite their large values. This is to prevent being selected as a representative color.
[0057]
The color replacement unit 15 selects the candidate color obtained in this way as a representative color, and is the representative color having the smallest Euclidean distance from the pixel color of the original image in the YC color space. This is a replacement unit that replaces the colors of the pixels of the original image. At that time, the number of pixels replaced with the representative color is counted for each representative color, and if the ratio of the count number to the count number of all the representative colors is equal to or less than a predetermined threshold value, the representative color is deleted. Do it. For the pixel that has been replaced with the deleted representative color, a new representative color having a shorter Euclidean distance is newly searched from the remaining representative colors, and replaced with the searched representative color.
[0058]
The storage unit 16 is a storage device such as a hard disk device, and stores image data 16a, parameter data 16b, and the like. Here, the image data 16a is image data such as a form, and the parameter data 16b is data storing threshold values used for various determinations. The control unit 17 is a control unit that performs overall control of the image color reduction device, and is a control unit that controls transmission and reception of various data between the functional units.
[0059]
Next, the hardware configuration of the image color reduction apparatus shown in FIG. 1 will be described. FIG. 12 is a diagram illustrating a hardware configuration of the image color reduction apparatus illustrated in FIG. 1. As shown in the figure, the image color reduction device has a configuration in which a keyboard 120, a display 121, a CPU 122, a RAM 123, an HDD 124, and a ROM 126 are connected by a bus 127.
[0060]
The hard disk (HD) 125, which is a storage medium that the HDD 124 stores and reads out, stores an image color reduction program 125a that is realized by executing the image color reduction method shown in the present embodiment with a computer. After being read, the program is analyzed by the CPU 122 and the image color reduction process is executed.
[0061]
This image color reduction process corresponds to the functions of the color space conversion unit 12, the two-dimensional histogram maximum value search unit 13, the one-dimensional histogram maximum value search unit 14, and the color replacement unit 15 shown in FIG. Image data 16a, parameter data 16b, and the like are also stored in the HD 125, read into the RAM 123, and referred to by the CPU 122.
[0062]
Next, a processing procedure of image color reduction processing performed by the image color reduction device according to the present embodiment will be described. FIG. 13 is a flowchart showing a processing procedure of image color reduction processing performed by the image color reduction device according to the present embodiment.
[0063]
As shown in the figure, first, based on the equations (1) to (5), the RGB values of the image data represented in the RGB color system are converted into YC values in the YC color system (step S1301). ). Then, a two-dimensional histogram is created from the three-dimensional histogram in the color space of the YC color system (step S1302).
[0064]
Then, the created two-dimensional histogram is smoothed (step S1303), and the maximum point of the smoothed two-dimensional histogram is searched (step S1304). Subsequently, a one-dimensional histogram is created on the basis of the found maximum point of the two-dimensional histogram, and a maximum point in the created one-dimensional histogram is searched (step S1305). This local maximum is a candidate for a representative color that replaces the color included in the image.
[0065]
After that, it is checked whether or not the smoothing process performed in step S1303 is the first smoothing process (step S1306). If it is the first smoothing process (step S1306, Yes), the process proceeds to step S1303. Then, the processing from step S1303 to step S1305 is performed again.
[0066]
If it is not the first smoothing process (step S1306, No), the color corresponding to the local maximum point searched when the previous smoothing process was performed and the current smoothing process are searched. Among the Euclidean distances in the YC color space between the colors corresponding to the maximum points, the largest one is extracted, and it is checked whether or not the maximum Euclidean distance is larger than a predetermined threshold (step) S1307). If it is not greater than the predetermined threshold value (No in step S1307), the process proceeds to step S1303, and the processes from step S1303 to step S1306 are performed again.
[0067]
If it is larger than the predetermined threshold value (step S1307, Yes), the color corresponding to the maximum point searched when the previous smoothing process was performed is selected as the representative color, and the similarity between these representative colors is selected. When the degrees are similar, the representative colors are integrated (step S1308). Thereafter, the color of each pixel included in the original image is replaced with a representative color (step S1309), and the image color reduction process is terminated.
[0068]
Next, the processing procedure of processing A shown in FIG. 13 will be described in detail. FIG. 14 is a flowchart showing a detailed processing procedure of the processing A shown in FIG. As shown in the figure, in the process A, first, a horizontal Sobel filter is applied to the value of the luminance Y of each pixel of the image converted into the YC color system, and the equation (7) is used. The value xbuf is calculated (step S1401). Similarly, a vertical Sobel filter is applied to the luminance Y value of each pixel, and the ybuf value is calculated using equation (8) (step S1402).
[0069]
Subsequently, the value of the edge strength w1 is calculated using Equation (6) (Step S1403), and the histogram addition value w is calculated from the edge strength w1 using Equation (9) (Step S1404). Then, a histogram addition value w of all pixels corresponding to the coordinate point (C1, C2, Y) in the color space is added to create a three-dimensional histogram, and the point having the same color difference coordinate (C1, C2) A process for creating a two-dimensional histogram in which the frequency values of the three-dimensional histogram are accumulated is performed (step S1405).
[0070]
Thereafter, it is checked whether or not the processing from step S1401 to step S1405 has been completed for all the pixels (step S1406). If there is a pixel that has not been processed yet (step S1406, No), the process proceeds to step S1401. Migrate and execute the process. If the processing has been completed for all pixels (step S1406, Yes), the processing A is terminated as it is.
[0071]
Next, the processing procedure of processing B shown in FIG. 13 will be described in detail. FIG. 15 is a flowchart showing a detailed processing procedure of the processing B shown in FIG. As shown in the figure, in the process B, first, the frequency values of the two-dimensional histograms at the eight neighboring points surrounding the color difference point (C1, C2) on the C1-C2 plane of interest are acquired (step S1501). .
[0072]
Then, it is checked whether the frequency value of the color difference point of interest is larger than the frequency value of the neighboring 8 points (step S1502). If the frequency value is not larger than the frequency value of the neighboring 8 points (step S1502) (S1502, No), the process proceeds to step S1505. If the frequency value is larger than the frequency values of the neighboring eight points (step S1502, Yes), the frequency value of the color difference point of interest is a value obtained by adding the frequency values of all points on the C1-C2 plane. It is checked whether or not it is 0.0001% or more (step S1503).
[0073]
If the frequency value is not 0.0001% or more (step S1503, No), the process proceeds to step S1505. If the frequency value is 0.0001% or more (step S1503, Yes), the color difference point (C1, C2) of interest is extracted as a local maximum point (step S1504).
[0074]
Then, it is checked whether or not the processing from step S1501 to step S1504 has been completed for all color difference points (step S1505). If there is a color difference point for which processing has not yet been completed (No in step S1505). ), The process proceeds to step S1501. When the process is completed for all the color difference points (step S1505, Yes), the process B is terminated as it is.
[0075]
Next, the processing procedure of processing C shown in FIG. 13 will be described in detail. FIG. 16 is a flowchart showing a detailed processing procedure of the process C shown in FIG. As shown in the figure, in process C, first, a one-dimensional histogram is generated for a straight line that passes through the maximum point in the two-dimensional histogram related to the C1-C2 plane and is perpendicular to the C1-C2 plane (step S1601). When creating a one-dimensional histogram, the value of the frequency corresponding to each point on the straight line is calculated using Equation (12) and Equation (13).
[0076]
Then, it is checked whether or not the frequency value of the point on the straight line of interest is larger than the frequency values of the points on both sides (step S1602). Step S1602, No), the process proceeds to step S1605. When the frequency value is larger than the frequency values of the points on both sides (step S1602, Yes), the frequency value of the target point is 0.001 which is a value obtained by adding the frequency values of all the points on the straight line. It is checked whether or not it is at least% (step S1603).
[0077]
When the frequency value is not 0.001% or more (step S1603, No), the process proceeds to step S1605. When the frequency value is 0.001% or more (step S1603, Yes), the point of interest is extracted as a maximum point (step S1604).
[0078]
Thereafter, it is checked whether or not the maximum point extraction processing shown in steps S1602 to S1604 has been completed for all points on the straight line (step S1605). If there is a point for which processing has not been completed (step S1605). (S1605, No), it moves to step S1602 and performs processing. If the processing has been completed for all points (step S1605, Yes), it is checked whether or not the processing from step S1601 to step S1605 has been completed for all maximum points of the two-dimensional histogram (step S1606).
[0079]
If the processing for all local maximum points has not been completed (step S1606, No), the processing proceeds to step S1601. When the process for all the local maximum points is completed (step S1606, Yes), the process C is terminated as it is.
[0080]
Next, the processing procedure of processing D shown in FIG. 13 will be described in detail. FIG. 17 is a flowchart showing a detailed processing procedure of the processing D shown in FIG. As shown in the figure, in process D, first, Euclidean distances for all sets of colors (representative candidate colors) corresponding to the maximum points of the one-dimensional histogram are calculated (step S1701).
[0081]
Then, an evaluation value for each set of colors is calculated by the equation (14) (step S1702). Subsequently, a set of colors having the smallest evaluation value is searched (step S1703), and it is checked whether or not the searched evaluation value is larger than a predetermined threshold value (step S1704).
[0082]
If the evaluation value is not larger than the threshold value (step S1704, No), the color with the smaller ratio is excluded by excluding the color with the smaller ratio of the frequency to the total frequency of the representative color candidate colors. The two colors are integrated (step S1705), and the process proceeds to step S1703. When the evaluation value is larger than the threshold value (step S1704, Yes), the process D is terminated as it is.
[0083]
Next, the processing procedure of processing E shown in FIG. 13 will be described in detail. FIG. 18 is a flowchart showing a detailed processing procedure of the process E shown in FIG. As shown in the figure, in the process E, first, for every other pixel of the image, the Euclidean distance between the representative color corresponding to the extracted maximum point and the color of the pixel is calculated in the YC color space. The pixel color is replaced with the minimum representative color (step S1801). Then, the number of pixels replaced for each representative color is counted (step S1802).
[0084]
Subsequently, it is checked whether or not the processing of step S1801 and step S1802 has been completed for every other pixel (step S1803). If the processing has not been completed for all pixels (step S1803). (S1803, No), the process proceeds to step S1801 to execute processing.
[0085]
If the processing is completed for all pixels (step S1803, Yes), whether the count number for each representative color is greater than 0.1% of the sum of the count numbers for all the representative colors Is checked (step S1804). If the count is not larger than 0.1% (No at Step S1804), the representative color that is not larger is excluded (Step S1806), and the process proceeds to Step S1805. As a result, representative colors with too few pixels for color replacement are excluded.
[0086]
If the count is larger than 0.1% (step S1804, Yes), it is checked whether or not the processing of step S1804 or step S1806 has been completed for all the representative colors (step S1805), and all the representative colors are checked. If the process has not been completed (No in step S1805), the process proceeds to step S1804 to execute the process.
[0087]
When the processing is completed for all the representative colors (step S1805, Yes), the pixel of interest is replaced with a pixel that has not been replaced with the representative color or with a representative color from which the representative color has been excluded. If the pixel is not replaced with a representative color or is not replaced with a excluded representative color (No at step S1807), the process proceeds to step S1809. .
[0088]
If the pixel is not replaced with a representative color or is a pixel replaced with an excluded representative color (step S1807, Yes), the YC color space between the pixel color and the color of the pixel is used. Is replaced with a representative color that minimizes the Euclidean distance (step S1808).
[0089]
Thereafter, it is checked whether or not the processing of step S1807 or step S1808 has been completed for all the pixels (step S1809). If there is a pixel that has not been processed yet (step S1809, No), the process proceeds to step S1807. Migrate and execute the process. If the processing has been completed for all pixels (step S1809, Yes), the processing E is terminated as it is.
[0090]
FIG. 19 is a diagram illustrating an example of an image processed by the image color reduction processing according to the present embodiment. FIG. 19A is a bitmap image represented by the RGB color system, and FIG. 19B is a diagram of the JPEG (Joint Photographic Experts Group) standard created from the bitmap image of FIG. 19A. FIG. 19C is an image of the JPEG2000 standard created from the bitmap image of FIG. 19A, and FIG. 19D is a diagram illustrating the image color reduction method described in this embodiment. The image is a PNG (Portable Network Graphics) standard image created from the 19 (a) bitmap image.
[0091]
As shown in FIG. 4D, the image color reduction method according to this embodiment not only has a high processing speed for creating an image, but also displays an image more clearly than an image in the JPEG standard or JPEG2000 standard. Therefore, it is particularly suitable when processing an image such as a form scanned with a scanner or the like at high speed.
[0092]
As described above, in the present embodiment, a two-dimensional histogram related to the C1-C2 plane is created based on the color distribution in the color space, and the maximum value of the created two-dimensional histogram is searched and searched. A one-dimensional histogram relating to a straight line that passes through a point corresponding to the determined local maximum value and is perpendicular to the C1-C2 plane is searched, the local maximum value of the generated one-dimensional histogram is searched, and the local maximum value of the searched one-dimensional histogram is Since the color corresponding to is selected as the representative color and the color included in the image is replaced with the selected representative color, the representative color is searched three-dimensionally in the color space of the YC color system. Instead, the representative color can be efficiently selected in a short processing time by dividing the search into two-dimensional search and one-dimensional search.
[0093]
In addition, since the smoothing process is appropriately performed on the two-dimensional histogram related to the C1-C2 plane, the representative color can be selected accurately.
[0094]
The embodiments of the present invention have been described so far, but the present invention is not limited to the above-described embodiments, and can be applied in various different embodiments within the scope of the technical idea described in the claims. It may be implemented.
[0095]
For example, in this embodiment, an RGB color system image is converted to a YC color system image, and the color-reduction processing is performed on the converted image. However, the present invention is not limited to this. Instead, the same processing is applied to the Lab color system image including the lightness L, the element a having a chromaticity ranging from green to red, and the element b having a chromaticity ranging from blue to yellow. be able to.
[0096]
In this embodiment, the two-dimensional histogram is smoothed to search for the maximum value of the smoothed two-dimensional histogram. However, the present invention is not limited to this, and smoothing is performed. The maximum point may be searched without performing the above, and the colors corresponding to the searched maximum points may be integrated and selected as the representative color. Thus, by not performing smoothing, the representative color can be selected in a shorter time.
[0097]
In addition, among the processes described in this embodiment, all or part of the processes described as being automatically performed can be manually performed, or the processes described as being manually performed All or a part of the above can be automatically performed by a known method. In addition, the processing procedure, control procedure, specific name, and information including various data and parameters shown in the above-described document and drawings can be arbitrarily changed unless otherwise specified.
[0098]
Each component of each illustrated device is functionally conceptual and does not necessarily need to be physically configured as illustrated. In other words, the specific form of distribution / integration of each device is not limited to that shown in the figure, and all or a part thereof may be functionally or physically distributed or arbitrarily distributed in arbitrary units according to various loads or usage conditions. Can be integrated and configured. Further, all or any part of each processing function performed in each device may be realized by a CPU and a program analyzed and executed by the CPU, or may be realized as hardware by wired logic.
[0099]
Note that the image color reduction method described in this embodiment can be realized by executing a program prepared in advance on a computer such as a personal computer or a workstation, as described in FIG. This program can be distributed via a network such as the Internet. The program can also be executed by being recorded on a computer-readable recording medium such as a hard disk, a flexible disk (FD), a CD-ROM, an MO, and a DVD and being read from the recording medium by the computer.
[0100]
【The invention's effect】
As described above, according to the first aspect of the present invention, the image in the color space includes the two-dimensional histogram relating to the predetermined plane of the color space in which the color characteristics included in the image are replaced with the three-dimensional coordinates. A one-dimensional histogram relating to a straight line that is created based on the color distribution, searches for the maximum value of the generated two-dimensional histogram, passes through a point corresponding to the searched maximum value, and is perpendicular to a predetermined plane of the color space Is created based on the distribution of colors contained in the image in the color space, the maximum value of the created one-dimensional histogram is searched, and the color corresponding to the searched maximum value of the one-dimensional histogram is selected as the representative color. However, since the color included in the image is replaced with the selected representative color, the representative color is not searched three-dimensionally in the color space, but is divided into a two-dimensional search and a one-dimensional search. Search By selecting a representative color can be created efficiently small images data size in a short processing time.
[0101]
According to the second aspect of the present invention, a two-dimensional histogram in consideration of the edge strength of the image is created based on the distribution of colors included in the image in the color space, and the maximum value of the created two-dimensional histogram is searched. Therefore, selecting the color of noise generated at the edge portion as the representative color can be efficiently suppressed.
[0102]
According to the invention of claim 3, the distribution of the color corresponding to the point that passes through the point corresponding to the maximum value and is within the predetermined distance of the color space from the point on the straight line perpendicular to the predetermined plane of the color space. Based on the above, a one-dimensional histogram related to a straight line perpendicular to a predetermined plane is created, and the maximum value of the created one-dimensional histogram is searched. Therefore, by examining the color distribution in the color space, A representative color showing a dense distribution can be extracted efficiently.
[0103]
According to the invention of claim 4, a two-dimensional histogram relating to a predetermined plane of the color space is created based on the distribution of colors included in the image in the color space, and the created two-dimensional histogram is smoothed. The maximum value of the smoothed two-dimensional histogram is searched, and a one-dimensional histogram relating to a straight line passing through a point corresponding to the searched maximum value and perpendicular to a predetermined plane is included in the image in the color space. Created based on the color distribution, searches for the maximum value of the created one-dimensional histogram, calculates the similarity between colors corresponding to the maximum value of the one-dimensional histogram before and after smoothing, and calculates Whether to further smooth the two-dimensional histogram, search for the maximum value of the smoothed two-dimensional histogram, and search for the maximum value of the one-dimensional histogram based on the similarity Therefore, the number of local maximum points to be extracted is controlled by performing smoothing and determining the number of times of smoothing based on the similarity between colors before and after smoothing, An appropriate number of representative colors can be obtained.
[0104]
According to the invention of claim 5, the colors corresponding to the plurality of maximum values of the searched one-dimensional histogram are integrated based on the similarity of the colors, and the integrated color is selected as the representative color. Since the colors included in the image are replaced with the representative colors, it is possible to efficiently create an image with a smaller data size by integrating similar colors.
[0105]
According to the invention of claim 6, a two-dimensional histogram relating to a predetermined plane of the color space in which the characteristics of the colors included in the image are replaced with three-dimensional coordinates is used as the distribution of colors included in the image in the color space. A search is made for a maximum value of the generated two-dimensional histogram, and a one-dimensional histogram relating to a straight line passing through a point corresponding to the searched maximum value and perpendicular to a predetermined plane of the color space is displayed in the color space. Is created based on the distribution of colors included in the image, and the local maximum value of the generated one-dimensional histogram is searched, and the color corresponding to the local maximum value of the searched one-dimensional histogram is selected as the representative color. Since the representative color is replaced with the color included in the image, instead of searching for the representative color three-dimensionally in the color space, the search is divided into a two-dimensional search and a one-dimensional search. Select representative color The Rukoto, it is possible to create an efficient small images data size in a short processing time.
[0106]
According to the invention of claim 7, a two-dimensional histogram taking into account the edge strength of the image is created based on the distribution of colors included in the image in the color space, and the maximum value of the created two-dimensional histogram is searched. Therefore, selecting the color of noise generated at the edge portion as the representative color can be efficiently suppressed.
[0107]
According to the eighth aspect of the present invention, the distribution of colors corresponding to points within a predetermined distance in the color space from points on a straight line passing through the point corresponding to the maximum value and perpendicular to the predetermined plane in the color space. Based on the above, a one-dimensional histogram related to a straight line perpendicular to a predetermined plane is created, and the maximum value of the created one-dimensional histogram is searched. Therefore, by examining the color distribution in the color space, A representative color showing a dense distribution can be extracted efficiently.
[0108]
According to the invention of claim 9, a two-dimensional histogram relating to a predetermined plane of the color space is created based on the distribution of colors included in the image in the color space, and the created two-dimensional histogram is smoothed. The maximum value of the smoothed two-dimensional histogram is searched, and a one-dimensional histogram relating to a straight line passing through a point corresponding to the searched maximum value and perpendicular to a predetermined plane is included in the image in the color space. Created based on the color distribution, searches for the maximum value of the created one-dimensional histogram, calculates the similarity between colors corresponding to the maximum value of the one-dimensional histogram before and after smoothing, and calculates Whether to further smooth the two-dimensional histogram, search for the maximum value of the smoothed two-dimensional histogram, and search for the maximum value of the one-dimensional histogram based on the similarity Therefore, the number of local maximum points to be extracted is controlled by performing smoothing and determining the number of times of smoothing based on the similarity between colors before and after smoothing, An appropriate number of representative colors can be obtained.
[0109]
According to the invention of claim 10, the colors corresponding to the plurality of maximum values of the searched one-dimensional histogram are integrated based on the similarity of the colors, and the integrated color is selected as the representative color. Since the colors included in the image are replaced with the representative colors, it is possible to efficiently create an image with a smaller data size by integrating similar colors.
[0110]
According to the invention of claim 11, a two-dimensional histogram relating to a predetermined plane of the color space in which the characteristics of the colors included in the image are replaced with three-dimensional coordinates is used as the distribution of colors included in the image in the color space. A search is made for a maximum value of the generated two-dimensional histogram, and a one-dimensional histogram relating to a straight line passing through a point corresponding to the searched maximum value and perpendicular to a predetermined plane of the color space is displayed in the color space. Is created on the basis of the distribution of colors included in the image in the image, the maximum value of the created one-dimensional histogram is searched, the color corresponding to the searched maximum value of the one-dimensional histogram is selected as the representative color, and selected. Since the representative color is replaced with the color included in the image, instead of searching for the representative color three-dimensionally in the color space, the search is divided into a two-dimensional search and a one-dimensional search. Select representative color By, it is possible to create an efficient small images data size in a short processing time.
[Brief description of the drawings]
FIG. 1 is a functional block diagram illustrating a configuration of an image color reduction apparatus according to an embodiment.
FIG. 2 is a diagram illustrating an example of an RGB image subjected to color reduction processing by the image color reduction device according to the present embodiment.
FIG. 3 is a diagram illustrating an example of a color space of image data obtained by converting an RGB color system to a YC color system.
FIG. 4 is a diagram illustrating an operator of a Sobel filter used for weighting each pixel.
FIG. 5 is a diagram illustrating an example of a two-dimensional histogram in a C1-C2 plane.
FIG. 6 is a diagram illustrating an operator of a smoothing filter applied to a two-dimensional histogram related to a C1-C2 plane.
FIG. 7 is a diagram illustrating an example of a straight line that passes through the maximum point of the two-dimensional histogram related to the C1-C2 plane and is perpendicular to the C1-C2 plane.
FIG. 8 is a conceptual diagram illustrating a method for creating a one-dimensional histogram.
FIG. 9 is a diagram showing a concept of a local maximum point in a one-dimensional histogram.
FIG. 10 is a diagram showing changes in the number of maximum points and the number of colors of representative color candidates by smoothing processing.
FIG. 11 is a diagram illustrating an example of a two-dimensional histogram on a C1-C2 plane that has been smoothed.
12 is a diagram showing a hardware configuration of the image color reduction apparatus shown in FIG. 1. FIG.
FIG. 13 is a flowchart showing a processing procedure of image color reduction processing performed by the image color reduction device according to the present embodiment.
14 is a flowchart showing a detailed processing procedure of processing A shown in FIG.
15 is a flowchart showing a detailed processing procedure of processing B shown in FIG.
16 is a flowchart showing a detailed processing procedure of processing C shown in FIG. 13;
FIG. 17 is a flowchart showing a detailed processing procedure of processing D shown in FIG. 13;
FIG. 18 is a flowchart showing a detailed processing procedure of processing E shown in FIG. 13;
FIG. 19 is a diagram showing an example of an image processed by the image color reduction processing according to the present embodiment.
[Explanation of symbols]
10 Input section
11 Display
12 color space converter
13 Two-dimensional histogram maximum value search unit
14 One-dimensional histogram maximum value search unit
15 color replacement part
16 Memory unit
16a, 123b Image data
16b, 123c Parameter data
120 keyboard
121 display
122 CPU
123 RAM
123a, 125a Image color reduction program
124 HDD
125 HD
126 ROM
127 bus

Claims (11)

An image color reduction device that selects a representative color representing a color included in an image, replaces the color included in the image with the representative color, and creates an image in which the number of colors of the image is reduced,
A two-dimensional histogram relating to a predetermined plane of a color space in which the characteristics of the colors included in the image are replaced with three-dimensional coordinates is created based on the distribution of colors included in the image in the color space. Two-dimensional histogram maximum value searching means for searching for a maximum value of the two-dimensional histogram;
A one-dimensional histogram relating to a straight line passing through a point corresponding to the maximum value searched by the two-dimensional histogram maximum value search means and perpendicular to a predetermined plane of the color space is displayed for the color included in the image in the color space. A one-dimensional histogram maximum value searching means for generating a maximum value of the generated one-dimensional histogram created based on the distribution;
Color replacement means for selecting a color corresponding to the maximum value of the one-dimensional histogram searched by the one-dimensional histogram maximum value search means as the representative color, and replacing the color included in the image with the selected representative color; An image color reduction device comprising: The two-dimensional histogram maximum value searching means generates the two-dimensional histogram in consideration of the edge strength of the image based on the distribution of colors included in the image in the color space, and the maximum value of the generated two-dimensional histogram The image color reduction device according to claim 1, wherein the image color reduction device is searched. The one-dimensional histogram maximum value searching means passes through a point corresponding to the maximum value searched by the two-dimensional histogram maximum value searching means, and starts from a point on a straight line perpendicular to a predetermined plane of the color space. A one-dimensional histogram relating to a straight line perpendicular to the predetermined plane is generated based on a color distribution corresponding to a point within a predetermined distance, and a maximum value of the generated one-dimensional histogram is searched. The image color reduction device according to claim 1 or 2. The two-dimensional histogram maximum value search means creates a two-dimensional histogram relating to a predetermined plane of the color space based on the distribution of colors included in the image in the color space, Smoothing is performed to search for a maximum value of the smoothed two-dimensional histogram, and the one-dimensional histogram maximum value searching means passes through a point corresponding to the maximum value searched by the two-dimensional histogram maximum value searching means. Generating a one-dimensional histogram relating to a straight line perpendicular to the predetermined plane based on a distribution of colors included in the image in the color space, searching for a maximum value of the generated one-dimensional histogram, and performing the smoothing The similarity between colors corresponding to the maximum value of the one-dimensional histogram before and after the smoothing is calculated, and the two-dimensional histogram maximum value search is performed based on the calculated similarity It is determined whether to further perform smoothing of the two-dimensional histogram by the stage, searching for the maximum value of the smoothed two-dimensional histogram, and searching for the maximum value of the one-dimensional histogram by the one-dimensional histogram maximum value searching means. The image color reduction device according to claim 1, 2, or 3. The color replacement unit integrates colors corresponding to a plurality of maximum values of the one-dimensional histogram searched by the one-dimensional histogram maximum value search unit based on the similarity of the colors, and the integrated color is the representative color. 5. The image color reduction device according to claim 1, wherein the color included in the image is replaced with the selected representative color. An image color reduction method for selecting a representative color representing a color included in an image, replacing the color included in the image with the representative color, and creating an image with a reduced number of colors in the image,
A two-dimensional histogram relating to a predetermined plane of a color space in which the characteristics of the colors included in the image are replaced with three-dimensional coordinates is created based on the distribution of colors included in the image in the color space. A two-dimensional histogram maximum value searching step of searching for a maximum value of the two-dimensional histogram;
A one-dimensional histogram relating to a straight line passing through a point corresponding to the maximum value searched in the two-dimensional histogram maximum value search step and perpendicular to a predetermined plane of the color space is displayed for the color included in the image in the color space. A one-dimensional histogram maximum value search step that is created based on the distribution and searches for the maximum value of the generated one-dimensional histogram;
A color replacement step of selecting a color corresponding to the maximum value of the one-dimensional histogram searched in the one-dimensional histogram maximum value search step as the representative color, and replacing a color included in the image with the selected representative color; An image color reduction method comprising: The two-dimensional histogram maximum value search step creates the two-dimensional histogram in consideration of the edge strength of the image based on the distribution of colors included in the image in the color space, and the maximum value of the generated two-dimensional histogram The image color reduction method according to claim 6, wherein the image color reduction method is searched. The one-dimensional histogram maximum value searching step passes through a point corresponding to the maximum value searched by the two-dimensional histogram maximum value searching step and starts from a point on a straight line perpendicular to a predetermined plane of the color space. A one-dimensional histogram relating to a straight line perpendicular to the predetermined plane is generated based on a color distribution corresponding to a point within a predetermined distance, and a maximum value of the generated one-dimensional histogram is searched. The image color reduction method according to claim 6 or 7. In the two-dimensional histogram maximum value searching step, a two-dimensional histogram relating to a predetermined plane of the color space is created based on a distribution of colors included in the image in the color space. Smoothing is performed to search for a maximum value of the smoothed two-dimensional histogram, and the one-dimensional histogram maximum value searching step passes through a point corresponding to the maximum value searched by the two-dimensional histogram maximum value searching step. Generating a one-dimensional histogram relating to a straight line perpendicular to the predetermined plane based on a distribution of colors included in the image in the color space, searching for a maximum value of the generated one-dimensional histogram, and performing the smoothing The similarity between colors corresponding to the maximum value of the one-dimensional histogram before and after the smoothing is calculated, and the two-dimensional histogram maximum value search is performed based on the calculated similarity It is determined whether or not to further perform smoothing of the two-dimensional histogram according to the process, search for the maximum value of the smoothed two-dimensional histogram, and search for the maximum value of the one-dimensional histogram by the one-dimensional histogram maximum value search step. The image color reduction method according to claim 6, 7 or 8. The color replacement step integrates colors corresponding to a plurality of maximum values of the one-dimensional histogram searched by the one-dimensional histogram maximum value search step based on the similarity of the colors, and the integrated color is the representative color 10. The image color reduction method according to claim 6, wherein a color included in the image is replaced with the selected representative color. An image color reduction program for selecting a representative color representing a color included in an image, replacing the color included in the image with the representative color, and creating an image with a reduced number of colors of the image,
A two-dimensional histogram relating to a predetermined plane of a color space in which the characteristics of the colors included in the image are replaced with three-dimensional coordinates is created based on the distribution of colors included in the image in the color space. A two-dimensional histogram maximum value search procedure for searching for a maximum value of a two-dimensional histogram;
A one-dimensional histogram relating to a straight line passing through a point corresponding to the maximum value searched by the two-dimensional histogram maximum value search procedure and perpendicular to a predetermined plane of the color space is displayed for the color included in the image in the color space. A one-dimensional histogram maximum value search procedure for creating a maximum value of the generated one-dimensional histogram created based on the distribution;
A color replacement procedure for selecting a color corresponding to the maximum value of the one-dimensional histogram searched by the one-dimensional histogram maximum value search procedure as the representative color, and replacing a color included in the image with the selected representative color; An image color reduction program for causing a computer to execute the above.

Images