JP4138126B2 - Image processing device - Google Patents

Description

図７は、係数発生部３０４の構成を示すものである。
【００５０】
係数発生部３０４は、ミクロ像域識別部１０４で生成した識別信号、及び制御部１１０より設定される拡大縮小率を示す値を入力し、画像の領域に応じたフィルタ係数を発生する。具体的には２５個のテーブルにより構成される。各テーブルはそれぞれ１０個の値を持ち、ＲＯＭ（読出し専用メモリ）により構成される。
【００５１】
これら１０個の値は、識別信号及び拡大縮小率Ｒにより選択される。すなわち、拡大縮小率が５０％以下、５０〜８０％、８０％〜１２５％、１２５％〜２００％、２００％以上の５通りと識別信号が０、１の２通りの計１０通りの場合に応じて、１０個のいずれかの値が選択される。すなわち、識別信号及び拡大縮小率を前記の５通りに類別した信号によりアドレスされるＲＯＭの内容が係数信号として出力される。このようなテーブル内容の例を図８に示す。
【００５２】
なお、本実施例ではテーブルの内容をすべてＲＯＭ（読出し専用メモリ）に格納しているが、ＲＡＭ（読み書き可能メモリ）に格納してもよい。この場合、拡大縮小率に応じてＲＡＭの内容を書き換えることにより、テーブルに格納するデータを２個に抑えることができ、メモリの容量を小さくすることができる。
【００５３】
また、原稿モードなどの他のモード指定によりフィルタ係数を切り替える場合も同じメモリ容量で実現できる。
【００５４】
なお、ここではＣ成分の画像信号に対するフィルタ３０１についてのみ説明したが、Ｍ成分の画像信号、Ｙ成分の画像信号もまったく同様の処理を行う。
【００５５】
次に、本実施例の動作と本発明を適用した効果について、従来例と比較しながら具体的に説明する。
【００５６】
本実施例の高画質化処理部１０５では畳み込み演算、いわゆるデジタルフィルタ処理を行っている。このフィルタ演算は識別信号及び拡大縮小率に応じて図８に示した係数発生部３０４のテーブルに従って切り替えられる。まず、拡大縮小率が１００％すなわち等倍の場合を考えると、識別結果が文字領域、網点領域、その他の領域で選択されるフィルタのカーネルを図９、その周波数応答特性を図１０に示す。ただし、フィルタは回転対称で縦横とも同じ周波数特性なので１次元の特性のみ表示する。
【００５７】
図１０からわかるように、本実施例での網点領域は１７５ｄｐｉの成分を除き、文字領域は１５０ｄｐｉの成分を強調し、その他領域では１２０ｄｐｉの成分をやや強調するような設計となっている。これは、通常のカラー印刷では１７５ｄｐｉの網点画像が多く用いられるのでこの周波数成分を除くこと、文字領域は細かい文字まで鮮鋭に再現するため高い周波数まで強調すること、それ以外の部分では画像入力部１０１でのぼけを補正することを目的にした設定となっている。ただし、これらは係数発生部３０４内のテーブルにより自由に設定できるので、画像入力部１０１の読取り特性、視覚特性や画質目標に応じて適当に設計が可能である。
【００５８】
これに対して従来例の高画質化処理について説明する。特公平６−５８８５号公報には図１１に示すとおり、画像にエッジ強調部１００３でエッジ強調処理と平滑部１００２で平滑化処理を施したのち、エッジ検出部１００１から出力されるエッジ検出信号で制御される混合比率により、これらの信号を混合部１００４で混合する方式が開示されている。エッジ強調部１００３、平滑部１００２は、それぞれ２次微分フィルタ、平均化フィルタにより構成されている。それらのフィルタのカーネルと周波数特性を図１２、図１３に示す。
【００５９】
このため、混合部１００４までの出力の周波数特性は、２つのフィルタの周波数特性の線形和となり、図１３の２つの周波数特性を線形混合したものとなる。この混合した例を破線で示す。このため、平滑化とエッジ強調の周波数特性は独立に設定可能であるが、それ以外の特性はこれらの線形和の範囲しか実現できない。このため、本実施例のように、文字領域、網点領域、その他領域でそれぞれ強調周波数や除去周波数を自由に設定することができず、すべての領域を目標とおりに高画質化することが困難となる。
【００６０】
また、本実施例では係数発生部３０４のテーブルにより拡大縮小率に応じたフィルタ係数を設定することができる。２００％拡大時のフィルタのカーネルは図１４に示すとおり縦横が非対称であり、縦方向のサイズの大きい構成となっている。前記でも述べたように、既存の多くのデジタル複写機と同様に、本実施例では画像入力部１０１で副走査方向のサンプリング密度を変えることにより、副走査方向の拡大縮小を行っている。さらに、高画質化処理部１０５の後段でデジタル処理により主走査方向の拡大縮小を行つているため、高画質化処理部１０５での画像信号は縦と横のサンプリング密度が異なる。
【００６１】
２００％拡大の場合、具体的には縦方向１２００ｄｐｉ、横方向６００ｄｐｉとなる。このため、本実施例のようにフィルタサイズを縦方向より横方向より大きくすることにより、最終的な出力画像の縦横の周波数特性はほぼ同じとなる。ただし、ここで縦とは副走査方向、横とは主走査方向を表わす。なお、拡大時は識別結果によりフィルタを切り替える効果が少ないので、本実施例では拡大時には識別信号によらずフィルタ係数を同じ設定としているが、必ずしもこれに限るものではなく、拡大時にも識別信号によりフィルタ係数を切り変える設定としてもよい。
【００６２】
一方、従来例では拡大縮小率に応じてフィルタのカーネルを変えることができないので、拡大縮小時にエッジ強調や平滑化の度合が方向により異なった信号が出力される。このため、線の方向により鮮鋭度の異なる不均一な画像が記録される。
【００６３】
以上説明したように上記第１実施例によれば、フィルタを構成する乗算係数を像域信号および拡大縮小倍率に応じて独立に変更することができる。これにより、信号処理部の周波数特性や縦横の方向性を画像の領域と拡大縮小率に応じて自由に変更することが可能となる。これにより、拡大時での画像のぼけや拡大縮小時での画像の縦横方向の解像度の不均一性を防ぐことができる。これらにより、より高画質な画像を再現する処理を提供することが可能となる。
【００６４】
次に、第１実施例の第１変形例について説明する。本変形例ではフィルタ部の構成が第１実施例と異なっている。
【００６５】
図１５は、この構成を示すものである。本変形例では、４個のラインディレイと２０個の画素ディレイにから出力された２５個の画像信号を生成するまでは第１実施例と同様であるが、これらの信号を加算器群１３０１に入力したのち、積和計算部１３０２に入力する。加算器群１３０１ではＸ軸、Ｙ軸に関して軸対称位置にある画像信号同士の加算を行う。具体的には下記式に示す加算を行う。なお、Ｑ８は加算ではないが表記の便宜上信号Ｑ８＝Ｐ２，２とする。
【００６６】
Ｑ０＝Ｐ０，０＋Ｐ０，４＋Ｐ４，０＋Ｐ４，４
Ｑ１＝Ｐ０，１＋Ｐ０，３＋Ｐ４，１＋Ｐ４，３
Ｑ２＝Ｐ０，２＋Ｐ４，２
Ｑ３＝Ｐ１，０＋Ｐ１，４＋Ｐ３，０＋Ｐ３，４
Ｑ４＝Ｐ１，１＋Ｐ１，３＋Ｐ３，１＋Ｐ３，３
Ｑ５＝Ｐ１，２＋Ｐ３，２
Ｑ６＝Ｐ２，０＋Ｐ２，４
Ｑ７＝Ｐ２，１＋Ｐ２，３
Ｑ８＝Ｐ２，２
本変形例の係数発生部１３０３では９個の係数を発生する。係数の発生にテーブルを用いる点、識別信号と拡大縮小率によりテーブル内の値が選択される点は、第１実施例と同様であるが発生する係数信号が９個である点が異なる。この発生した係数信号と加算器群１３０１で生成した９個の画像信号とを積和計算部１３０２で積和演算し、この結果が出力信号となる。
【００６７】
本変形例では、通常のフィルタの係数が縦および横の軸について軸対称であることを利用して、軸対称な位置の画像信号を予め加算してから共通な係数を乗算する。本変形例の係数テーブルの内容を図１６に示す。第１実施例の係数テーブルと等価な処理になるように構成されている。
【００６８】
以上説明したように上記第１実施例の第１変形例によれば、第１実施例の効果に加え、係数発生部のテーブルサイズが小さくなること、積和計算部での乗算演算数が減ることにより回路規模を小さくすることができる。なお、その分フィルタ係数の自由度が減るが、通常の条件ではフィルタのカーネルは縦横とも軸対称であり実用上なんら問題はない。
【００６９】
次に、第１実施例の第２変形例について説明する。
【００７０】
第１実施例では識別信号は文字領域、網点領域、その他領域の３種類をとったが、本変形例では文字領域識別とその他領域の識別の間に２レべル、および網点領域識別とその他領域の識別の間に２レベル設け、計７種類の識別信号を設ける。すなわち、ミクロ像域識別部１０４で、文字領域とその他領域の中間的な性質をもつ領域についてはその性質の度合に応じて、中間信号値を出力し、また網点領域とその他領域の中間的な性質をもつ領域についても同様に中間信号を出力する。識別信号の値とその意味を図１７に示す。すなわち、識別信号０が網点領域、識別信号１がかなり網点領域、識別信号２がやや網点領域、識別信号３がその他領域、識別信号４がやや文字領域、識別信号５がかなり文字領域、識別信号６が文字領域としている。
【００７１】
本変形例の係数発生部では、識別信号の種類（本例では７種類）に応じた係数テーブルを用意しておく。中間信号に応じた係数は中間的な値としておくことで、中間的な識別信号に対しては中間的な処理を行うことができる。
【００７２】
以上説明したように上記第１実施例の第２変形例によれば、中間値の信号値を用いることにより上記で示したように係数発生部のテーブルは大きくなるが、一方、識別誤りによる画質劣化を小さくすることと、２つの領域の境界で処理が大きく変化することによる弊害を防ぐことができるという利点がある。
【００７３】
次に、第２実施例について説明する。
【００７４】
図１８は、この発明の画像処理装置を適用した第２実施例のデジタルカラー複写機の構成を示すものである。この装置は、画像入力部１４０１、色変換部１４０２、マクロ像域識別部１４０３、ミクロ像域識別部１４０４、高画質化処理部１４０５、拡大縮小部１４０６、墨入れ処理部１４０７、階調処理部１４０８、画像記録部１４０９、及びこれら各部の制御などを行う制御部１４１０より構成されている。本実施例では、マクロ像域識別部１４０３、ミクロ像域識別部１４０４、高画質化処理部１４０５の構成、及びその間の信号の流れが第１実施例と異なっている。それ以外の画像入力部１４０１、色変換部１４０２、拡大縮小部１４０６、墨入れ処理部１４０７、階調処理部１４０８、画像記録部１４０９は第１実施例とまったく同様であるので説明を省略する。
【００７５】
以下、異なる部分について構成と動作を詳しく説明する。
【００７６】
本実施例のマクロ像域識別部１４０３の構成は、第１実施例と同様であるが、プログラムＲＯＭにしたがつて実行するので領域分析の内容と識別信号の種類が若干異なっている。マクロ像域識別部１４０３は、第１実施例の識別動作に加え、粗い網点領域と細かい網点領域の識別を行う。ここで網点の粗さは、１５０ｄｐｉを目安として、これより周波数が高い領域を細かい、周波数が低い領域を粗い網点と定義する。これらの分離は、ソフト処理で画像の周期成分を調べる処理を追加することにより可能である。
【００７７】
図１９は、第２の実施例におけるマクロ識別信号の値と領域の関係を示すものである。すなわち、信号値０が通常文字領域、信号値１が背景上文字領域、信号値２が連続階調領域、信号値３が細かい網点階調領域、信号値４が粗い網点階調領域、信号値５がその他領域としている。
【００７８】
図２０は、ミクロ像域識別部１４０４の構成を示すものである。ミクロ像域識別部１４０４は、像域識別部１５０１，１５０２，１５０３から構成されている。このミクロ像域識別部１４０４では、ＹＭＣの色ごとに独立して識別処理を行い、色ごとに独立した識別信号を出力する。
【００７９】
まず、色変換部１４０２から出力された画像信号のＣＭＹのそれぞれの色成分であるＣ成分の画像信号１５５１、Ｍ成分の画像信号１５５２、Ｙ成分の画像信号１５５３は、それぞれの色の像域識別部１５０１〜１５０３に入力する。各像域識別部（１５０１〜１５０３）には、マクロ識別信号１５５４も入力する。各像域識別部（１５０１〜１５０３）では、各色成分の画像信号とマクロ像域信号とを用いてそれぞれに領域識別を行い、文字、網点、それ以外の３領域に分類し、それぞれ値２、０、１をとるミクロ識別信号１５５５、１５５６、１５５７を出力する。ミクロ識別信号（１５５５〜１５５７）は、該当する色成分の文字や網点があるかどうかの識別を行う。
【００８０】
例えば、黒色はＹＭＣのすべての成分をもち、緑色はＣとＹの成分を、赤色はＭとＹの成分のみをもつので、Ｍ成分のミクロ識別信号１５５６は黒文字や赤文字の領域は文字領域として識別され、緑文字領域はその他の領域と識別される。同様に、黒や赤の網点部分は網点領域として識別されるが、緑色の網点の領域はその他領域として識別される。
【００８１】
図２１は、高画質化処理部１４０５の構成を示すものである。高画質化処理部１４０５は、遅延回路１６０１、加算器群１６０２、第１の積和演算部１６０３、第１の係数発生部１６０４、第２の積和演算部１６０５、第２の係数発生部１６０６、及び加算部１６０８とから構成されている。なお、第１実施例と同様に高画質化処理部１４０５は、Ｃ，Ｍ，Ｙの３色の色信号ごとに３チャネルの同じ処理により構成されているので、ここではＣ信号の処理部のみ説明する。
【００８２】
まず、色変換部１４０２から出力されたＣ成分の画像信号１６５１は、ライン遅延と画素遅延よりなる遅延回路部１６０１で遅延され、注目画素周辺の縦横各５×５画素の２５個の信号を並列に出力する。なお、この遅延回路部１６０１は、第１実施例と同様なので説明は省略する。第１実施例と同様にｍラインの遅延とｎ画素の遅延を受けた画像信号をＰｍ，ｎと表記する。遅延回路部１６０１から出力された２５個の並列画像信号Ｐｍ，ｎは、加算器群１６０２に入力され上述した加算処理を施され、縮退した画像信号Ｑ０〜Ｑ８が出力される。
【００８３】
縮退画像信号Ｑ０〜Ｑ８は、第１の積和演算部１６０３及び第２の積和演算部１６０５に入力する。第１の積和演算部１６０３では、第１の係数発生部１６０４より出力する９個の第１の係数信号Ａ０〜Ａ８と縮退画像信号との間で下記に示す積和演算処理を行い、その演算結果の信号Ｒ０が出力される。
【００８４】
Ｒ０＝ΣＡｉ−Ｑｉ
ｉ
第１の係数発生部１６０４は、ＲＡＭより構成され、ここに９個の係数が格納されている。このＲＡＭの内容が、そのまま第１の係数信号として出力される。９個の係数を図２２に示す。これらの係数により第１の積和演算部１６０３では平均化フィルタ処理が実現される。
【００８５】
ＲＡＭの内容は、１ページの画像信号を処理している間は固定であるが、その他の期間は制御部１４１０から設定することができる。このため、例えば拡大縮小時や設定される原稿モードによってこの係数値を書き換えることができる。
【００８６】
第２の積和演算部１６０５では、第２の係数発生部１６０６より出力する９個の第２の係数信号Ｂ０〜Ｂ８と縮退画像信号Ｑ０〜Ｑ８の間で下記式に示す積和演算処理を行ない、その演算結果の信号Ｒ１が出力される。
【００８７】

図２３は、第２の係数発生部１６０６の構成を示すものである。この第２の係数発生部１６０６は、識別変換部１７０１と係数テーブル１７０２とから構成されている。第２の係数発生部１６０６では、ミクロ像域識別部１４０４で生成したＣ色の識別信号、マクロ像域識別部１４０３から出力するマクロ識別信号を入力し、画像の領域に応じた係数を発生する。
【００８８】
識別変換部１７０１には、マクロ識別信号とＣ色のミクロ識別信号が入力される。識別変換部１７０１は、ハードウエア的にはルックアップテーブルにより構成されている。前記で説明したようにマク口識別信号は６通りの値を、ミクロ識別信号は３通りの値をとるが、識別変換部１７０１はルックアップテーブルにより３ビット（ｂｉｔ）の係数選択信号を出力する。
【００８９】
識別変換部１７０１から出力した３ビット（ｂｉｔ）の係数選択信号により、係数テーブル１７０２の内容を選択する。係数テーブル１７０２は、９個の係数の組が８種類格納されており、係数選択信号によりこの８種類の中から選択された信号が出力される。これら２つのテーブルではマクロ識別信号とミクロ識別信号により決まる画像領域に応じて最適な係数が選択される。ここで、マクロ識別信号とミクロ識別信号はそれぞれ６通りおよび３通りで計１８通りの組合わせがある。
【００９０】
これらの信号は独立ではないのでこの１８通りには冗長な分類がある。このため識別変換部１７０１のテーブルにより５通りに縮退させ、係数テーブル１７０２の容量を小さくすることができる。この６通りの信号の意味は０：粗い網点、１：細かい網点、２：背景上文字、３：文字、４：階調上のエッジ、５：その他としている。識別変換部１７０１のルックアップテーブルの入力と出力信号の関係を図２４に、係数テーブル１７０２の内容を図２５に示す。
【００９１】
図２５に示されるように、第２の積和演算部１６０５では領域に応じた周波数特性の２次微分フィルタ演算が行われる。具体的には粗い網点部では１２０ｄｐｉの周波数成分を除き、細かい網点部では１７５ｄｐｉの周波数成分を除き、背景文字では１２０ｄｐｉの周波数成分を除き、文字領域では１５０ｄｐｉの周波数成分を強調し、階調上のエッジ領域では強調し、１２０ｄｐｉの周波数成分をやや強調その他の領域ではフラットな周波数特性とする。
【００９２】
最後に加算部１６０８で第１の積和演算部１６０３の出力信号Ｒ０と第２の積和演算部１６０５の出力信号Ｒ１に対して下記式の演算を行ない、高画質化処理部の出力信号Ｒ２として出力する。
【００９３】
Ｒ２＝Ｒ０＋ｆ（Ｒ１）
ここでｆ（）は非線形変換関数であり、その入出力特性を図２６に示す。この変換はエッジ強調による画像エッジ部での過強調を防ぐためのものである。
【００９４】
以上はＣ信号について説明したが、Ｍ信号、Ｙ信号についても同様の処理を行う。ただし、入力する画像信号はそれぞれＭ成分、Ｃ成分の画像信号を用いる。
【００９５】
以上説明したように上記第２実施例によれば、第１の積和演算部で固定の平均化フィルタ処理、第２の積和演算部で２次微分フィルタ処理を行い、２次微分処理の係数を識別信号に応じて切り替えることにより、第１実施例と同様に全体のフィルタ特性を自由に切替えることができる。
【００９６】
また、２次微分処理出力は画像の高域周波数成分、平均化処理出力は画像の低域周波数成分を表わし、これらを分離することによりフィルタ係数の設計が容易になる。
【００９７】
また、低域成分は画像の階調再現性に対して大きく寄与するのに対し、高域成分は画像の鮮鋭さに寄与するため、高域成分だけに非線形変換処理を施すことにより、階調再現性をそこなうことなく鮮鋭度を改善することができる。
【００９８】
また、本実施例ではマクロ像域識別により識別された結果により直接フィルタ係数を選択できる。例えば、本実施例で粗い網点および細かい網点の場合に選択されるフィルタのカーネルをそれぞれ図２８の（ａ）と（ｂ）に示す。細かい網点では外周部の係数が０となっており、等価的に３×３とサイズの小さいフィルタが選択されることになる。一方、粗い網点では５×５のフィルタが選択され、網点の粗さにより等価的にフィルタサイズすなわちカットオフ周波数を切り替えることが可能となる。このように粗い網点ではカットオフ周波数を下げることによりモアレの発生を防ぎ、細かい網点領域ではカットオフ周波数を上げることにより画像の鮮鋭感を保つことができる。
【００９９】
また、本実施例ではＣＭＹの成分ごとに係数発生部をもつので、ＣＭＹごとに独立にフィルタ係数の設定が可能である。
【０１００】
画像入力部では一般に光学レンズにより原稿をＣＣＤ上に結像させるが、レンズの色収差によりＲＧＢ成分でＭＴＦ特性を同一にすることは難しい。また、３ラインタイプのＣＣＤセンサを用いる場合、センサ間の読取り位置の補正を行う際に、拡大縮小率によっては補間演算を行なう必要があるが、この場合にも副走査方向の解像度が低下する。
【０１０１】
ＲＧＢのうち１つの色成分の解像度が低い場合、色変換後のＣＭＹ信号ともに解像度に影響を受けるが、その補色への影響が最も大きい。例えば、Ｂ成分の信号のＭＴＦ特性が低い場合、Ｂの補色であるＹ信号のＭＴＦ特性が最も影響を受けて低くなり、記録画像の画質としてはＹ成分のある赤や緑の文字の鮮鋭度が悪くなる。本実施例ではフィルタの係数をＣＭＹごとにフィルタ係数を適正値に設定することにより、この影響を補正し、どの色の文字も同様に精細に再現することが可能となる。
【０１０２】
次に、第２実施例の変形例について説明する。本変形例は第２実施例に対し、第２の係数発生部の構成だけが異なっている。そこで、この部分についてのみ説明する。
【０１０３】
図２７は、本変形例に係る第２の係数発生部１６０７の構成を示すものである。第２の係数発生部１６０７は、識別変換テーブル１９０１、乗数テーブル１９０２、係数テーブル１９０３、及び乗算器群１９０４よりなる。
【０１０４】
識別変換テーブル１９０１には、マクロ識別信号とミクロ識別信号を入力し、これらの値の組によってアドレスされる３ｂｉｔの乗数選択信号が出力される。
【０１０５】
乗数テーブル１９０２には、８個の乗数値が格納されており、乗数選択信号により８個のいずれかが選択され、乗数信号として出力される。
【０１０６】
係数テーブル１９０３には、９個の係数が格納されており、これらの値は常に９個の係数信号として出力され、乗算器群１９０４内の９個の乗算器でそれぞれ乗数信号と乗算され、係数信号として出力され、第２の積和演算部１６０５に供給される。
【０１０７】
本変形例では、識別信号に応じて選択された乗数信号により係数信号が生成され、これによりフィルタの特性を制御することができる。１自由度の乗数信号で制御を行うため、フィルタの周波数特性の自由度は小さくなるが、一方係数テーブルや乗数テーブルなどのテーブルの総サイズを小さくすることができる。
【０１０８】
また、第２の積和演算処理後に非線形変換などを行ってから加算を行うため、高域成分を低域成分と独立して処理が可能である。
【０１０９】
なお、本発明の適用例として、フィルタ処理の係数の切替えの例を中心に説明したが、本発明はこれに限るものではない。例えば、識別信号に応じて色変換処理の処理パラメータや墨入れ処理部の墨入れ係数を切り替えることも可能である。これにより、画像の領域に応じて記録色度を色再現性重視や濃度重視などの間で切り替えることが可能となる。
【０１１０】
以上説明したように上記発明の実施の形態によれば、フィルタ処理などの画像処理の特性を画像の領域に応じてきめ細かく制御することができ、細かい文字の精細再現、銀塩写真などの階調画像の鮮鋭再現、網点でのモアレの発生の防止などを同時に実現することができる。
【０１１１】
【発明の効果】
以上詳述したようにこの発明によれば、フィルタ処理の自由度を上げ、細かい網点と粗い網点、細かい文字と太い文字など多くの領域種や拡大縮小時の縦横の画像の周波数特性の違いに応じて信号処理の周波数特性を自由に変調できる画像処理装置を提供することができる。
【図面の簡単な説明】
【図１】この発明の画像処理装置を適用した第１実施例のデジタルカラー複写機の構成を示すブロック図。
【図２】マクロ像域識別部の構成例を示す図。
【図３】領域分離信号の値と領域の関係を説明するための図。
【図４】高画質化処理部の構成を示す図。
【図５】第１のフィルタ部の構成を示す図。
【図６】遅延回路部の構成を示す図。
【図７】係数発生部の構成を示す図。
【図８】係数発生部のテーブルの内容を示す図。
【図９】各領域のフィルタカーネルを示す図。
【図１０】各領域の周波数応答特性を示す図。
【図１１】従来の高画質化処理を説明するための図。
【図１２】従来のフィルタカーネルを示す図。
【図１３】従来のフィルタの周波数特性を示す図。
【図１４】２００％拡大時のフィルタのカーネルを示す図。
【図１５】第１実施例の第１変形例のフィルタ部の構成を示す図。
【図１６】係数発生部のテーブルの内容を示す図。
【図１７】係数発生部からの識別信号を説明するための図。
【図１８】この発明の画像処理装置を適用した第２実施例のデジタルカラー複写機の構成を示すブロック図。
【図１９】マクロ識別信号の値と領域の関係を説明するための図。
【図２０】ミクロ像域識別部の構成を示す図。
【図２１】高画質化処理部の構成を示す図。
【図２２】第１の係数発生部の係数を示す図。
【図２３】第２の係数発生部の構成を示す図。
【図２４】識別変換部のルックアップテーブルの入力と出力信号の関係を示す図。
【図２５】係数テーブルの内容を説明するための図。
【図２６】加算部の非線形変換の特性を示す図。
【図２７】変形例の第２の係数発生部の構成を示す図。
【図２８】粗い網点および細かい網点の場合に選択されるフィルタのカーネルを示す図。
【符号の説明】
１０１，１４０１…画像入力部（画像入力手段）
１０２，１４０２…色変換部
１０３，１４０３…マクロ像域識別部（識別手段）
１０４，１４０４…ミクロ像域識別部（識別手段、出力手段）
１０５，１４０５…高画質化処理部
１０６，１４０６…拡大縮小部
１０７，１４０７…墨入れ処理部
１０８，１４０８…階調処理部
１０９，１４０９…画像記録部
１１０，１４１０…制御部
３０１，３０２，３０３…フィルタ部（演算手段）
３０４…係数発生部（発生手段）
１５０１，１５０２，１５０３…像域識別部
１６０１…遅延回路部
１６０２…加算器群
１６０３…第１の積和演算部（第１の演算手段）
１６０４…第１の係数発生部
１６０５…第２の積和演算部（第２の演算手段）
１６０６…第２の係数発生部（発生手段）
１６０８…加算部（加算手段）[0001]
BACKGROUND OF THE INVENTION
The present invention relates to an image processing apparatus for improving the image quality of an electronic image by sharpening or smoothing.
[0002]
[Prior art]
In recent years, with the development of digital technology, the digitization of copiers is rapidly progressing. By digitizing an image signal, there are advantages such as integration with a printer function and a FAX function, and image quality improvement by applying a digital signal processing technique. In a digital copying machine, an original image is scanned by a scanner, read as an electrical signal, converted into a digital signal by an A / D converter, and then subjected to digital signal processing such as image area identification processing, filter processing, and gradation processing. For example, a recorded image is formed on output paper by an electrophotographic method or the like.
[0003]
Such a digital copying machine requires high image quality processing. In the scanner, since an original image is formed on a CCD by an optical system, a component having a high spatial frequency is deteriorated depending on the MTF characteristic of the optical system, and a phenomenon such as blurring occurs. This leads to resolution degradation of characters and line drawings. Further, in a general printed document, halftone expression is performed by a halftone method, and a periodic halftone pattern is formed. If this is processed as it is, there is a possibility that the halftone dot pattern becomes noise and the image quality deteriorates, or moire specific to digital processing occurs in the gradation processing section. For this reason, in order to record a high-quality image, it is necessary to apply an edge emphasis filter to the character or line drawing portion to correct the influence of the blur caused by the scanner. In a halftone printed document, it is necessary to apply a smoothing filter to remove halftone components.
[0004]
In order to selectively perform these processes, a method of performing image area identification and switching signal processing according to the identification result is conventionally known. In this method, first, an area type of a document is automatically identified from an image signal by image area identification processing. Specifically, it is identified as an area such as a character, halftone dot, photograph, or background. Then, the signal processing method is switched according to the identification signal. For example, an edge enhancement processing method is used in an area identified as a character, and a smoothing processing method is applied in an area identified as a halftone dot.
[0005]
As such image area identification means, for example, a method is known that uses a local density change difference between a gradation area and a character area or a local pattern difference. As an example of the former, Japanese Patent Application Laid-Open No. 58-3374 discloses that an image is divided into small blocks, and the difference between the maximum density and the minimum density in each block is obtained. If it is smaller, a method for identifying the block as a gradation image area is disclosed.
[0006]
As an example of the latter, Japanese Patent Application Laid-Open No. 60-204177 discloses a method of performing binarization after applying a Laplacian filter to an image, for example, for discrimination by the shape of a 4 × 4 pixel pattern. Yes.
[0007]
In the method of identifying by referring to these minute areas, there is a problem that a sharp area on a halftone dot area or a gradation image is easily misidentified as a character part. To solve this problem, the entire image is scanned with coarse resolution, and the structure of the image is analyzed by software to identify the image area (character area extraction algorithm for mixed mode communication, IEICE) The journal J67-D, v0111.pp1277-1284 (1984)), a method combining these methods, and the like have also been proposed. Each has advantages and disadvantages in performance and cost, and a suitable method is applied according to the required performance and application of the apparatus.
[0008]
As a signal processing switching method according to such an identification result, a method of switching between edge enhancement and smoothing is often used. When edge enhancement and smoothing are switched, image quality degradation is significant when noise occurs in the transition region between the character region and the other region, or misidentification occurs in the image region identification unit. A method of switching in stages is used. Japanese Examined Patent Publication No. 6-5858 discloses a method in which an edge enhancement process and a smoothing process are performed on an image and then these signals are mixed by a mixing ratio controlled by an edge detection signal output from an edge detection unit. It is disclosed.
[0009]
In Japanese Patent Laid-Open No. 60-167574, an edge component extraction process and a smoothing process are performed on an image signal, and then a signal obtained by multiplying the edge component signal by a coefficient controlled by an identification signal is added to the smoothing signal. A scheme is disclosed. Although these systems have slightly different configurations, as a result, the intensity of the edge component is modulated in accordance with the identification signal.
[0010]
However, in the prior art, the intensity of the edge component is modulated according to the identification result, so that the degree of freedom of process switching is low. For example, in the embodiment of the above reference (Japanese Patent Publication No. 6-5885), the gain of the intensity of the emphasized frequency can be changed, but the emphasized frequency itself cannot be fixed and changed. For this reason, it is impossible to change the cut-off frequency when performing smoothing. In addition, since the frequency characteristics in the main scanning direction and the sub-scanning direction are also fixed, the edge emphasis intensity cannot be modulated independently for the main scanning and the sub-scanning. In order to increase the degree of freedom of image processing in the prior art, it is possible to extend the embodiment of the reference to provide more than types of filters and modulate the mixing ratio of these outputs, but the processing circuit according to the degree of freedom Therefore, there is a problem that the circuit scale increases.
[0011]
As described above, when there are few identification area types such as a character area and two types other than the character area when applied to a copying machine, the low degree of freedom of image processing is not a big problem. However, recently, it has become possible to discriminate more finely and with many types of regions by using an identification means that utilizes the structure of the entire image as described above. For example, discrimination between coarse and fine halftone dots and discrimination between small characters and thick characters are possible. In this case, since the frequency characteristics of the image vary depending on the period of the halftone dots and the thickness of the characters, the frequency band to be smoothed and the frequency band to be emphasized differ depending on the region. Therefore, in order to realize further higher image quality, it is necessary to finely change the filter characteristics, that is, the frequency band and the gain, based on the identification result.
[0012]
In the case of enlargement / reduction in a digital copying machine, the sub-scanning direction is performed by changing the scanning speed of the scanner, and the main scanning direction is performed by signal processing. For this reason, when the filter process is provided before the enlargement / reduction by the signal process, the aspect ratio of the signal in the filter unit varies depending on the enlargement / reduction ratio. For this reason, it is necessary to change the vertical and horizontal frequency characteristics in accordance with the enlargement / reduction ratio.
[0013]
As described above, the conventional technology cannot deal with fine differences in image area types and cannot cope with differences in the vertical and horizontal characteristics of signals at the time of enlargement / reduction.
[0014]
[Problems to be solved by the invention]
As described above, in the conventional technique, the strength of the edge component is modulated according to the identification result, so that the degree of freedom of switching of processing is low. Recently, as described above, by using an identification means that utilizes the structure of the entire image, it has become possible to perform finer and more detailed identification of region types. For example, it is possible to distinguish between coarse and fine halftone dots. Discrimination between small characters and thick characters. In this case, the frequency characteristics of the image vary depending on the dot period and the thickness of the characters, so the frequency band to be smoothed and the frequency band to be emphasized differ depending on the region. In order to achieve high image quality, it is necessary to finely change the filter characteristics, that is, the frequency band and gain according to the identification result.
[0015]
In the case of enlargement / reduction in a digital copying machine, the sub-scanning direction is performed by changing the scanning speed of the scanner, and the main scanning direction is performed by signal processing. For this reason, when the filter process is provided before the enlargement / reduction by the signal process, the aspect ratio of the signal in the filter unit varies depending on the enlargement / reduction ratio. For this reason, it is necessary to change the vertical and horizontal frequency characteristics in accordance with the enlargement / reduction ratio.
[0016]
As described above, there is a problem that it is impossible to cope with a fine difference in image area type and a difference in vertical and horizontal characteristics of a signal at the time of enlargement / reduction.
[0017]
Therefore, the present invention increases the degree of freedom of filter processing, and performs signal processing according to the difference in frequency characteristics of vertical and horizontal images during enlargement / reduction and many types of regions such as fine halftone dots and coarse halftone dots, fine letters and thick letters. An object of the present invention is to provide an image processing apparatus that can freely modulate the frequency characteristics of the image.
[0018]
[Means for Solving the Problems]
The present invention provides a first calculation means for separating an input image signal based on density difference and saturation of surrounding pixels and halftone dot area roughness, and density distribution and density change from the input image signal. Extracting a certain piece of structural information, referring to the image data separated by the first computing means, a second computing means for separating each color component into a character area, a halftone dot area, and other areas; Output from the first computing means Separation signal And output from the second computing means Separation signal Based on , Record all pixels in front Pixel modulation processing Or, odd-numbered pixels are recorded in front, and even-numbered pixels are recorded in back. And a third calculation means for selecting two-pixel modulation processing, wherein the third calculation means includes a predetermined fixed coefficient, and the fixed coefficient And degenerate image signal First product-sum operation unit for averaging using And before Based on the respective outputs of the first calculation means and the second calculation means Product-sum operation processing is performed using the obtained second coefficient and the degenerate image signal. A second product-sum operation unit; An adder that adds and outputs the results calculated by the first and second product-sum operation units An image processing apparatus characterized by comprising:
[0022]
DETAILED DESCRIPTION OF THE INVENTION
Hereinafter, an embodiment of the present invention will be described with reference to the drawings.
[0023]
FIG. 1 shows the configuration of a digital color copying machine according to a first embodiment to which the image processing apparatus of the present invention is applied. This apparatus includes an image input unit 101, a color conversion unit 102, a macro image area identification unit 103, a micro image area identification unit 104, an image quality enhancement processing unit 105, an enlargement / reduction unit 106, an inking process unit 107, and a gradation processing unit. 108, an image recording unit 109, and a control unit 110 that controls these units. The image processing apparatus of the present invention is applied to the image quality enhancement processing unit 105. Note that descriptions of trimming, masking, and other editing processes not directly related to the present invention are omitted.
[0024]
First, the overall configuration will be briefly described.
[0025]
The control unit 110 includes a CPU, a RAM, a program ROM, and the like. The operation mode and calculation parameters of each processing unit are set according to information such as a document mode and an enlargement / reduction ratio instructed by the user through a control panel (not shown) of the copying machine.
[0026]
The image input unit 101 reads a document image and outputs a color image signal 151. The configuration of the image input unit 101 will be briefly described. A linear light source such as a halogen lamp irradiates light to a linear area on the document surface. Then, the linear region of the document surface is imaged on the color line sensor by a reduction optical system composed of three movable mirrors and lenses. The color line sensor is composed of three CCD line sensors each having R, G, and B color filters arranged in parallel and close to each other. The light quantity distribution on the CCD light receiving surface corresponding to the image information is converted into an electric signal. The signals are converted and sequentially read out, amplified by an amplifier, and then converted into a digital signal by an A / D converter. Further, there are a shading correction circuit for correcting CCD sensitivity unevenness and light source illuminance unevenness, and a delay / interpolation circuit for correcting differences in the positions of the three line sensors of RGB, but detailed description thereof is omitted.
[0027]
By repeating reading while scanning the movable mirror in a direction perpendicular to the line sensor, information on the entire surface of the original image is sequentially read. Here, the direction of the line sensor is called a main scanning direction, and the moving direction of the movable mirror is called a sub-scanning direction. When the image is enlarged or reduced, the scanning speed of the movable mirror in the sub-scanning direction is made inversely proportional to the enlargement / reduction magnification. Since the reading speed of the electric signal by the CCD is kept constant, the image signal is enlarged / reduced in the sub-scanning direction and the same magnification image signal is obtained in the main scanning direction.
[0028]
Note that the sampling density of the image input unit 101 in this embodiment is 600 dpi in both vertical and horizontal directions, that is, 600 pixels per inch.
[0029]
Next, the color conversion unit 102 converts the color image signal 151 representing the RGB reflectance into a YMC density signal 152 representing the density of the color material to be recorded. The relationship between the RGB reflectance and the YMC density is generally a complex non-linear relationship. Therefore, in order to realize this conversion process, a three-dimensional table lookup method or a method combining a one-dimensional table lookup and a 3 × 3 matrix is used. Specific configurations are described in detail in, for example, Japanese Patent Publication No. 1-055245 and Japanese Patent Publication No. Sho 61-007774.
[0030]
The macro image area identification unit 103 analyzes the structure of the entire image from the input image signal 151, identifies the image area, and outputs a macro image area signal.
[0031]
FIG. 2 shows a configuration example of the macro image area identification unit 103. The macro image area identification unit 103 includes an image separation unit 201, an image memory unit 202, a CPU 203, a program memory 204, and an area signal output unit 205. That is, the image separation unit 201 separates the color image signal 151 output from the image input unit 151 into image data of a plurality of planes according to the density difference and saturation of neighboring pixels, and sequentially stores them in the image memory unit 202. I will do it. The image memory unit 202 has a capacity corresponding to the number of image planes, and stores all separated image signals for one screen.
[0032]
Next, in accordance with the program code stored in the program memory (ROM) 204, the CPU 203 refers to the contents of the separated image data stored in the image memory unit 202, and separates the regions, and the separation result is displayed in the image memory unit. Write to 202. The area information stored in the image memory unit 202 is read by the area signal output unit 205 in synchronization with the second read signal from the image input unit 101. At this time, since the pixel density of the region separation information in the image memory unit 202 and the pixel density of the image signal from the image input unit 101 are different, the density conversion of the region separation signal is performed, and the pixel densities of both are matched and output. .
[0033]
In this embodiment, the region separation signal is represented by a 3-bit signal, and the relationship between the value and the region is as shown in FIG. That is, the signal value 0 represents a normal character region, the signal value 1 represents a background character region, the signal value 2 represents a continuous tone region, the signal value 3 represents a halftone region, and the signal value 4 represents another region. In addition, the region separation signal may be represented by a 5-bit signal, and each bit signal may represent five regions.
[0034]
The micro image area identification unit 104 extracts micro structure information such as density distribution and density change from the input image (density) signal 152, and further refers to the macro identification signal output from the macro area identification unit 103. , Separated into character area, halftone dot area, and other areas. Here, by switching the micro discrimination threshold and the discrimination process using the macro discrimination signal, the region having a pattern in which the micro structure information is similar to the character region, such as the edge of the gradation image, is correctly identified. be able to. Then, an identification signal 153 having a value of 2 in the character area, a value of 0 in the halftone area, and a value of 1 in the other areas is output.
[0035]
The character region is a region including the edge of the character or line drawing and the vicinity thereof. The other area is a non-halftone image such as a silver halide photograph or a background area where no characters are written. A specific method of image area identification is described in detail in Japanese Patent Laid-Open Nos. 2-199588 and 1-3783. By using the macro area identification signal, the macro structure information of the image can be used, so that the identification accuracy is improved.
[0036]
The high image quality processing unit 105 performs sharpening and smoothing processing on the YMC color image signal. In a copying machine, a document image is mainly used as a manuscript. In such an image, character images and gradation images are mixed. It is important that the character image is reproduced sharply. On the other hand, it is important that gradation images are reproduced smoothly. Also, printing and commercially available printers often use halftone dots for gradation expression, and it is important to remove this halftone component.
[0037]
For this reason, the image quality improvement processing unit 105 selectively switches the characteristics of the filter processing according to the image area signal 153 output from the macro image area identification unit 103 and the micro image area identification unit 104. When the image area signal represents a character, edge enhancement is performed by applying an edge enhancement filter. When the image area signal represents a gradation edge, a weak edge enhancement filter is applied. In addition, when the image area signal represents a gentle gradation, a smoothing filter is applied to remove noise and halftone components. Thereby, the character image can be reproduced sharply and the gradation image can be reproduced smoothly.
[0038]
The present invention is applied to the high image quality processing unit 105, and details will be described later.
[0039]
The enlargement / reduction unit 106 performs enlargement / reduction in the main scanning direction. As an enlargement / reduction processing method, a linear interpolation method, a projection method, or the like is known. When enlarging and reducing by digital signal processing, there are problems such as jaggy noise in which lines are jagged at the time of enlargement and moire generated in a halftone image at the time of reduction. The linear interpolation method hardly generates jaggy noise, and the projection method hardly generates moire noise. For this reason, in this embodiment, generation of these noises is suppressed by using a linear interpolation method at the time of enlargement and a projection method at the time of reduction. The linear interpolation method and the projection method are described in detail in Japanese Patent Application Laid-Open No. 2-308378. As described above, in the present embodiment, the image input unit 101 performs enlargement / reduction in the sub-scanning direction, and these can realize enlargement / reduction in both two-dimensional directions.
[0040]
The inking unit 107 converts the filtered color YMC signal into a YMCK 4th version signal. Although black can be expressed by overlapping the three color materials of YMC, the black color material generally has a higher density than the YMC layer and is cheaper. Recording is performed with the four colors of YMCK included. Specifically, methods such as UCR (Under Color Reduction) and GCR (Gray Component Removal) are known and are actually used. The calculation formula of the GCR method is shown below. However, the input CMY density signal is described as CMY, and the output CMYK density signal is described as C ′, M ′, Y ′, and K ′.
K ′ = k · min (C, M, Y)
C ′ = (C−K) / (1−K)
M ′ = (M−K) / (1−K)
Y ′ = (Y−K) / (1−K)
In recording such as electrophotography, the on / off length of light is modulated to express an intermediate density. The gradation processing unit 108 performs this modulation processing. Specifically, a pulse signal having a width corresponding to the density signal is generated. The on / off of the laser beam is controlled according to this pulse. Here, it is configured to be able to switch between the control for moving the pulse forward and the control for moving it back in the unit period.
[0041]
There are two modulation methods: 2-pixel modulation and 1-pixel modulation. In the two-pixel modulation method, the odd-numbered pixels are recorded in a front-aligned manner, and the even-numbered pixels are recorded in a back-aligned manner. On the other hand, in the one-pixel modulation method, all the pixels are recorded in a front-aligned manner. In the one-pixel modulation method, since the pulse ON / OFF cycle is in units of one pixel, recording can be performed with a resolution in units of one pixel. On the other hand, since the two-pixel modulation method has a period of two pixels, the resolution is lower than that of the one-pixel modulation method. However, since the pulse width for expressing the same density is doubled, the stability of density is increased and the gradation is improved compared to the one-pixel modulation method. Thus, the one-pixel modulation method is a method suitable for recording a character image, while the two-pixel modulation method is a method suitable for recording a gradation image.
[0042]
In this embodiment, 2-pixel modulation processing and 1-pixel modulation processing are selected according to the image area signal. Specifically, one-pixel modulation processing is selected when the image area signal represents a character, and two-pixel modulation processing is selected when the image area signal represents a tone edge or a gentle portion. As a result, an image having a smooth gradation and rich gradation can be reproduced in the gradation area, and a high-resolution and sharp image can be recorded in the character area.
[0043]
Next, the image recording unit 109 will be described. In this embodiment, an electrophotographic method is used for the image recording unit 109. The principle of the electrophotographic system will be briefly described. First, the intensity of laser light or the like is modulated in accordance with the image density signal, and this modulated light is irradiated onto the photosensitive drum. Electric charges corresponding to the amount of irradiation light are generated on the photosensitive surface of the photosensitive drum. Accordingly, a laser beam is scanned in the axial direction of the photosensitive drum in accordance with the scanning position of the image signal, and a two-dimensional charge distribution corresponding to the image signal is formed on the photosensitive drum by rotating and scanning the photosensitive drum. . Subsequently, the toner charged by the developing device is adhered on the photosensitive drum. At this time, an amount of toner corresponding to the potential adheres to form an image. The toner on the photosensitive drum is transferred onto a recording sheet via a transfer belt, and finally the toner is melted and fixed on the recording sheet by a fixing device. A full color image can be recorded on the paper surface by sequentially performing this operation for the four colors of YMCK.
[0044]
Next, the configuration and operation of the high image quality processing unit 105 to which the present invention is applied will be described in detail. The configuration of the image quality enhancement processing unit 105 is shown in FIG. The image quality improvement processing unit 105 includes first, second, and third filter units 301, 302, and 303 and a coefficient generation unit 304.
[0045]
The image signal 152 output from the color conversion unit 102 is input to the first, second, and third filter units 301, 302, and 303. The filter units 301, 302, and 303 respectively correspond to the CMY components of the image signal, the C component image signal 351 is filtered by the filter unit 301 to output the image signal 153 c, and the M component image signal 352 is output. The filter unit 302 performs filtering to output an image signal 153m, and the Y component image signal 353 is filtered by a filtering unit 303 to output an image signal 153y. Since the configurations of these filters are exactly the same, the first filter unit 301 will be described in detail below.
[0046]
FIG. 5 shows a configuration of the first filter unit 301. The filter unit 301 includes a delay circuit unit 401 and a product-sum operation unit 402. First, the input image signal 451 is delayed by the delay circuit unit 401, and 25 pixel signals 452 to 476 of 5 × 5 pixels in the vertical and horizontal directions around the target pixel are output in parallel. As shown in FIG. 6, the delay circuit unit 401 includes four line delays 901 to 904 and 20 pixel delays 905 to 924. As a result, image signals 951 and 961 to 964 from delay 0 line to delay 4 line are generated.
[0047]
These five image signals are further delayed by four pixel delays, respectively, and image signals delayed from four to four pixels are generated, for a total of 25 image signals. Here, an image signal that has received a delay of m lines and a delay of n pixels is represented as Pm, n (0 ≦ m <5, 0 ≦ n <5, m and n are integers).
[0048]
These image signals are input to the product-sum operation unit 402, and a product sum Q with the 25 coefficient signals 452 output from the coefficient generation unit 304 is calculated to be an output signal 453. This calculation formula is shown below.
[0049]

FIG. 7 shows the configuration of the coefficient generator 304.
[0050]
The coefficient generation unit 304 receives the identification signal generated by the micro image area identification unit 104 and the value indicating the enlargement / reduction ratio set by the control unit 110, and generates a filter coefficient corresponding to the region of the image. Specifically, it is composed of 25 tables. Each table has 10 values and is composed of a ROM (read only memory).
[0051]
These ten values are selected by the identification signal and the enlargement / reduction ratio R. That is, when the scaling ratio is 50% or less, 50 to 80%, 80% to 125%, 125% to 200%, 200% or more, and the identification signal is 0, 1 or 2 in 10 ways. In response, any one of 10 values is selected. That is, the contents of the ROM addressed by the signals classified into the above-described five classification signals and the enlargement / reduction ratio are output as coefficient signals. An example of such table contents is shown in FIG.
[0052]
In this embodiment, all the contents of the table are stored in ROM (read only memory), but may be stored in RAM (read / write memory). In this case, by rewriting the contents of the RAM according to the enlargement / reduction ratio, the data stored in the table can be reduced to two, and the capacity of the memory can be reduced.
[0053]
Further, the same memory capacity can be realized when the filter coefficient is switched by specifying another mode such as a document mode.
[0054]
Although only the filter 301 for the C component image signal has been described here, the same processing is performed for the M component image signal and the Y component image signal.
[0055]
Next, regarding the operation of the present embodiment and the effects of applying the present invention, Conventional A specific explanation will be given while comparing with examples.
[0056]
The high image quality processing unit 105 of this embodiment performs a convolution operation, so-called digital filter processing. This filter operation is switched according to the table of the coefficient generator 304 shown in FIG. 8 according to the identification signal and the enlargement / reduction ratio. First, considering the case where the enlargement / reduction ratio is 100%, that is, the same magnification, FIG. 9 shows a filter kernel selected in the character region, the halftone dot region, and other regions, and FIG. 10 shows the frequency response characteristics thereof. . However, since the filter is rotationally symmetric and has the same frequency characteristics both vertically and horizontally, only one-dimensional characteristics are displayed.
[0057]
As can be seen from FIG. 10, the halftone dot region in this embodiment is designed to emphasize the 150 dpi component except for the 175 dpi component, and slightly emphasize the 120 dpi component in the other regions. This is because, in normal color printing, a halftone dot image of 175 dpi is often used, so this frequency component is removed, the character area is sharply reproduced to a fine character, and high frequency is emphasized. The setting is intended to correct blur in the unit 101. However, since these can be freely set by a table in the coefficient generation unit 304, it is possible to design appropriately according to the reading characteristics, visual characteristics, and image quality target of the image input unit 101.
[0058]
On the other hand, the conventional high image quality processing will be described. In Japanese Examined Patent Publication No. 6-5858, as shown in FIG. 11, an edge emphasis unit 1003 applies an edge emphasis process to an image and a smoothing unit 1002 performs a smoothing process. Apply After that, a method is disclosed in which the mixing unit 1004 mixes these signals according to the mixing ratio controlled by the edge detection signal output from the edge detection unit 1001. The edge emphasizing unit 1003 and the smoothing unit 1002 are configured by a secondary differential filter and an averaging filter, respectively. The kernel and frequency characteristics of these filters are shown in FIGS.
[0059]
Therefore, the frequency characteristic of the output up to the mixing unit 1004 is a linear sum of the frequency characteristics of the two filters, and is a linear mixture of the two frequency characteristics shown in FIG. This mixed example is indicated by a broken line. For this reason, the frequency characteristics of smoothing and edge enhancement can be set independently, but other characteristics can be realized only in the range of these linear sums. For this reason, unlike the present embodiment, it is not possible to freely set the emphasis frequency and the removal frequency in the character area, halftone dot area, and other areas, respectively, and it is difficult to improve the image quality as intended in all areas. It becomes.
[0060]
In this embodiment, the filter coefficient corresponding to the enlargement / reduction ratio can be set by the table of the coefficient generator 304. As shown in FIG. 14, the filter kernel at 200% magnification is asymmetric in length and width and has a large size in the vertical direction. As described above, as in many existing digital copying machines, in this embodiment, the image input unit 101 changes the sampling density in the sub-scanning direction to perform enlargement / reduction in the sub-scanning direction. Further, since the enlargement / reduction in the main scanning direction is performed by digital processing at the subsequent stage of the image quality improvement processing unit 105, the vertical and horizontal sampling densities of the image signal in the image quality improvement processing unit 105 are different.
[0061]
In the case of 200% enlargement, specifically, the vertical direction is 1200 dpi and the horizontal direction is 600 dpi. Therefore, by making the filter size larger than the horizontal direction rather than the vertical direction as in this embodiment, the vertical and horizontal frequency characteristics of the final output image become substantially the same. Here, the vertical represents the sub-scanning direction, and the horizontal represents the main scanning direction. In addition, since there is little effect of switching the filter depending on the identification result at the time of enlargement, in this embodiment, the filter coefficient is set to be the same regardless of the identification signal at the time of enlargement. It is good also as a setting which switches a filter coefficient.
[0062]
On the other hand, in the conventional example, the kernel of the filter cannot be changed according to the enlargement / reduction ratio. Therefore, signals with different edge enhancement and smoothing degrees depending on directions are output during enlargement / reduction. For this reason, nonuniform images having different sharpness depending on the direction of the line are recorded.
[0063]
As described above, according to the first embodiment, the multiplication coefficient constituting the filter can be changed independently according to the image area signal and the enlargement / reduction ratio. As a result, the frequency characteristics and vertical and horizontal directions of the signal processing unit can be freely changed according to the image area and the enlargement / reduction ratio. Thereby, it is possible to prevent blurring of the image at the time of enlargement and nonuniformity of the resolution in the vertical and horizontal directions at the time of enlargement / reduction. Accordingly, it is possible to provide a process for reproducing a higher quality image.
[0064]
Next, a first modification of the first embodiment will be described. In this modification, the configuration of the filter unit is different from that of the first embodiment.
[0065]
FIG. 15 shows this configuration. This modification is the same as in the first embodiment until 25 image signals output from 4 line delays and 20 pixel delays are generated, but these signals are added to the adder group 1301. After input, it is input to the product-sum calculation unit 1302. The adder group 1301 performs addition of image signals at axially symmetric positions with respect to the X axis and the Y axis. Specifically, the addition shown in the following equation is performed. Although Q8 is not an addition, it is assumed that signal Q8 = P2, 2 for convenience of description.
[0066]
Q0 = P0,0 + P0,4 + P4,0 + P4,4
Q1 = P0, 1 + P0, 3 + P4, 1 + P4, 3
Q2 = P0,2 + P4,2
Q3 = P1,0 + P1,4 + P3,0 + P3,4
Q4 = P1,1 + P1,3 + P3,1 + P3,3
Q5 = P1,2 + P3,2
Q6 = P2,0 + P2,4
Q7 = P2,1 + P2,3
Q8 = P2,2
The coefficient generator 1303 of this modification generates nine coefficients. The point that the table is used for generating the coefficient and the value in the table is selected by the identification signal and the enlargement / reduction ratio are the same as in the first embodiment, except that nine coefficient signals are generated. The generated coefficient signal and the nine image signals generated by the adder group 1301 are subjected to a product-sum operation by a product-sum calculation unit 1302, and the result is an output signal.
[0067]
In this modification, by utilizing the fact that the coefficients of a normal filter are axially symmetric with respect to the vertical and horizontal axes, image signals at axially symmetric positions are added in advance and then multiplied by a common coefficient. The contents of the coefficient table of this modification are shown in FIG. The processing is equivalent to the coefficient table of the first embodiment.
[0068]
As described above, according to the first modification of the first embodiment, in addition to the effects of the first embodiment, the table size of the coefficient generator is reduced, and the number of multiplication operations in the product-sum calculator is reduced. As a result, the circuit scale can be reduced. Although the degree of freedom of the filter coefficient is reduced accordingly, under normal conditions, the filter kernel is axially symmetric both vertically and horizontally, and there is no practical problem.
[0069]
Next, a second modification of the first embodiment will be described.
[0070]
In the first embodiment, the identification signal has three types of character area, halftone dot area, and other area. However, in this modification, two levels between the character area identification and the other area identification, and the halftone dot area identification. Two levels are provided between the identification of the other areas and a total of seven types of identification signals. That is, the micro image area identification unit 104 outputs an intermediate signal value for an area having an intermediate property between the character area and the other area, and outputs an intermediate signal value between the halftone area and the other area. Similarly, an intermediate signal is output for a region having a characteristic. The value of the identification signal and its meaning are shown in FIG. That is, the identification signal 0 is a halftone dot region, the identification signal 1 is a considerable halftone dot region, the identification signal 2 is a little halftone dot region, the identification signal 3 is another region, the identification signal 4 is a little character region, and the identification signal 5 is a considerably character region. The identification signal 6 is a character area.
[0071]
In the coefficient generator of this modification, a coefficient table corresponding to the type of identification signal (seven types in this example) is prepared. By setting the coefficient corresponding to the intermediate signal to an intermediate value, intermediate processing can be performed on the intermediate identification signal.
[0072]
As described above, according to the second modified example of the first embodiment, the table of the coefficient generation unit is enlarged as described above by using the intermediate signal value, but on the other hand, the image quality due to the identification error is increased. There are advantages in that deterioration can be reduced and adverse effects caused by large changes in processing at the boundary between two regions can be prevented.
[0073]
Next, a second embodiment will be described.
[0074]
FIG. 18 shows the configuration of a digital color copying machine according to a second embodiment to which the image processing apparatus of the present invention is applied. This apparatus includes an image input unit 1401, a color conversion unit 1402, a macro image area identification unit 1403, a micro image area identification unit 1404, an image quality improvement processing unit 1405, an enlargement / reduction unit 1406, an inking processing unit 1407, and a gradation processing unit. 1408, an image recording unit 1409, and a control unit 1410 that controls these units. In the present embodiment, the configurations of the macro image area identification unit 1403, the micro image area identification unit 1404, and the image quality enhancement processing unit 1405, and the signal flow therebetween are different from those in the first example. Other image input unit 1401, color conversion unit 1402, enlargement / reduction unit 1406, inking process unit 1407, gradation processing unit 1408, and image recording unit 1409 are the same as those in the first embodiment, and thus description thereof is omitted.
[0075]
Hereinafter, the configuration and operation of different parts will be described in detail.
[0076]
The configuration of the macro image area discriminating unit 1403 of the present embodiment is the same as that of the first embodiment, but since it is executed according to the program ROM, the contents of the area analysis and the type of identification signal are slightly different. In addition to the identification operation of the first embodiment, the macro image area identification unit 1403 identifies a coarse halftone dot area and a fine halftone dot area. Here, with regard to the roughness of halftone dots, 150 dpi is used as a guide, and a region with a higher frequency than this is defined as fine, and a region with a low frequency is defined as a rough halftone dot. These separations are possible by adding a process for examining the periodic component of the image by a software process.
[0077]
FIG. 19 shows the relationship between the value of the macro identification signal and the area in the second embodiment. That is, the signal value 0 is a normal character area, the signal value 1 is a background character area, the signal value 2 is a continuous tone area, the signal value 3 is a fine halftone area, and the signal value 4 is a coarse halftone area. The signal value 5 is the other area.
[0078]
FIG. 20 shows the configuration of the micro image area identification unit 1404. The micro image area identification unit 1404 includes image area identification units 1501, 1502, and 1503. The micro image area identification unit 1404 performs an identification process independently for each color of YMC, and outputs an identification signal independent for each color.
[0079]
First, the C component image signal 1551, the M component image signal 1552, and the Y component image signal 1553, which are the respective CMY color components of the image signal output from the color conversion unit 1402, are image area identifications of the respective colors. Input to parts 1501 to 1503. A macro identification signal 1554 is also input to each image area identification unit (1501-1503). Each image area identification unit (1501-1503) performs area identification using the image signal of each color component and the macro image area signal, classifies them into three areas of characters, halftone dots, and other areas, each having a value of 2 , 0, 1 are output as micro identification signals 1555, 1556, 1557. The micro identification signal (1555-1557) identifies whether there is a character or halftone dot of the corresponding color component.
[0080]
For example, black has all the components of YMC, green has C and Y components, and red has only M and Y components, so the M component micro-identification signal 1556 is the character region for black and red characters. And the green character area is identified as other areas. Similarly, black and red halftone dots are identified as halftone areas, while green halftone areas are identified as other areas.
[0081]
FIG. 21 shows the configuration of the image quality enhancement processing unit 1405. The image quality improvement processing unit 1405 includes a delay circuit 1601, an adder group 1602, a first product-sum operation unit 1603, a first coefficient generation unit 1604, a second product-sum operation unit 1605, and a second coefficient generation unit 1606. , And an adder 1608. As in the first embodiment, the image quality improvement processing unit 1405 is configured by the same processing of three channels for each of the three color signals of C, M, and Y, and only the C signal processing unit is here. explain.
[0082]
First, the C component image signal 1651 output from the color conversion unit 1402 is delayed by a delay circuit unit 1601 composed of a line delay and a pixel delay, and 25 signals of 5 × 5 pixels in the vertical and horizontal directions around the target pixel are paralleled. Output to. Since the delay circuit unit 1601 is the same as that of the first embodiment, the description thereof is omitted. As in the first embodiment, an image signal that has received a delay of m lines and a delay of n pixels is represented as Pm, n. The 25 parallel image signals Pm, n output from the delay circuit unit 1601 are input to the adder group 1602 and subjected to the addition process described above, and degenerated image signals Q0 to Q8 are output.
[0083]
The degenerated image signals Q0 to Q8 are input to the first product-sum operation unit 1603 and the second product-sum operation unit 1605. The first product-sum operation unit 1603 performs the following product-sum operation processing between the nine first coefficient signals A0 to A8 output from the first coefficient generation unit 1604 and the degenerated image signal. An operation result signal R0 is output.
[0084]
R0 = ΣAi−Qi
i
The first coefficient generator 1604 is composed of a RAM, in which nine coefficients are stored. The content of this RAM is output as it is as the first coefficient signal. Nine coefficients are shown in FIG. With these coefficients, the first product-sum calculation unit 1603 realizes averaging filter processing.
[0085]
The content of the RAM is fixed while the image signal of one page is processed, but can be set from the control unit 1410 during other periods. Therefore, for example, the coefficient value can be rewritten at the time of enlargement / reduction or the set original mode.
[0086]
The second product-sum operation unit 1605 performs the product-sum operation processing shown in the following equation between the nine second coefficient signals B0 to B8 and the degenerated image signals Q0 to Q8 output from the second coefficient generation unit 1606. Then, a signal R1 as a result of the calculation is output.
[0087]

FIG. 23 shows the configuration of the second coefficient generator 1606. The second coefficient generation unit 1606 includes an identification conversion unit 1701 and a coefficient table 1702. The second coefficient generation unit 1606 receives the C color identification signal generated by the micro image area identification unit 1404 and the macro identification signal output from the macro image area identification unit 1403, and generates a coefficient corresponding to the area of the image. .
[0088]
The identification conversion unit 1701 receives a macro identification signal and a C-color micro identification signal. The identification conversion unit 1701 is configured by a lookup table in terms of hardware. As described above, the Mac mouth identification signal has six values and the micro identification signal has three values, but the identification conversion unit 1701 outputs a 3-bit (bit) coefficient selection signal using a lookup table. .
[0089]
The content of the coefficient table 1702 is selected by a 3-bit coefficient selection signal output from the identification conversion unit 1701. The coefficient table 1702 stores eight sets of nine coefficients, and a signal selected from the eight types is output by a coefficient selection signal. In these two tables, the optimum coefficient is selected according to the image area determined by the macro identification signal and the micro identification signal. Here, the macro identification signal and the micro identification signal are 6 kinds and 3 kinds, respectively, for a total of 18 combinations.
[0090]
Since these signals are not independent, there are redundant classifications in the 18 ways. For this reason, it is possible to reduce the capacity of the coefficient table 1702 by reducing the capacity of the coefficient table 1702 by using the table of the identification conversion unit 1701. The meanings of these six signals are 0: coarse halftone dot, 1: fine halftone dot, 2: character on background, 3: character, 4: edge on gradation, 5: other. FIG. 24 shows the relationship between the input and output signals of the lookup table of the identification conversion unit 1701, and FIG. 25 shows the contents of the coefficient table 1702.
[0091]
As shown in FIG. 25, the second product-sum operation unit 1605 performs a second-order differential filter operation with a frequency characteristic corresponding to the region. Specifically, the rough halftone dot portion excludes the 120 dpi frequency component, the fine halftone dot portion excludes the 175 dpi frequency component, the background character excludes the 120 dpi frequency component, and the character region emphasizes the 150 dpi frequency component. The enhancement edge region is emphasized, and the frequency component of 120 dpi has a flat frequency characteristic in the other enhancement region.
[0092]
Finally, the adder 1608 performs the following expression on the output signal R0 of the first product-sum operation unit 1603 and the output signal R1 of the second product-sum operation unit 1605, and outputs the output signal R2 of the image quality improvement processing unit. Output as.
[0093]
R2 = R0 + f (R1)
Here, f () is a nonlinear conversion function, and its input / output characteristics are shown in FIG. This conversion is to prevent over-emphasis at the image edge portion due to edge enhancement.
[0094]
Although the C signal has been described above, the same processing is performed for the M signal and the Y signal. However, the input image signals are M component and C component image signals, respectively.
[0095]
As described above, according to the second embodiment, the first product-sum operation unit performs fixed averaging filter processing, and the second product-sum operation unit performs secondary differential filter processing. By switching the coefficient according to the identification signal, the entire filter characteristics can be freely switched as in the first embodiment.
[0096]
The secondary differential processing output represents the high frequency component of the image, and the averaging processing output represents the low frequency component of the image. By separating these, the filter coefficient can be easily designed.
[0097]
In addition, the low-frequency component greatly contributes to the gradation reproducibility of the image, whereas the high-frequency component contributes to the sharpness of the image. Sharpness can be improved without losing reproducibility.
[0098]
In this embodiment, the filter coefficient can be directly selected based on the result identified by the macro image area identification. For example, FIG. 28A and FIG. 28B respectively show the filter kernels selected in the case of coarse halftone dots and fine halftone dots in this embodiment. For fine halftone dots, the coefficient of the outer peripheral portion is 0, and a filter having a small size of 3 × 3 is equivalently selected. On the other hand, a 5 × 5 filter is selected for coarse halftone dots, and the filter size, that is, the cutoff frequency can be switched equivalently depending on the roughness of the halftone dots. In such a rough halftone dot, the occurrence of moire can be prevented by lowering the cut-off frequency, and in the fine halftone dot region, the sharpness of the image can be maintained by raising the cut-off frequency.
[0099]
In addition, in the present embodiment, since a coefficient generation unit is provided for each CMY component, filter coefficients can be set independently for each CMY.
[0100]
In an image input unit, an original is generally imaged on a CCD by an optical lens, but it is difficult to make the MTF characteristics the same for RGB components due to chromatic aberration of the lens. When a 3-line type CCD sensor is used, it is necessary to perform an interpolation operation depending on the enlargement / reduction ratio when correcting the reading position between the sensors. In this case, however, the resolution in the sub-scanning direction is lowered. .
[0101]
When the resolution of one color component of RGB is low, both the CMY signals after color conversion are affected by the resolution, but the influence on the complementary color is the largest. For example, when the MTF characteristic of the B component signal is low, the MTF characteristic of the Y signal that is the complementary color of B is most affected and becomes low, and the image quality of the recorded image is the sharpness of red and green characters having the Y component. Becomes worse. In this embodiment, by setting the filter coefficient to an appropriate value for each CMY, the influence can be corrected and characters of any color can be reproduced precisely in the same manner.
[0102]
Next, a modification of the second embodiment will be described. The present modification is different from the second embodiment only in the configuration of the second coefficient generator. Therefore, only this part will be described.
[0103]
FIG. 27 shows a configuration of the second coefficient generator 1607 according to the present modification. The second coefficient generation unit 1607 includes an identification conversion table 1901, a multiplier table 1902, a coefficient table 1903, and a multiplier group 1904.
[0104]
The identification conversion table 1901 receives a macro identification signal and a micro identification signal, and outputs a 3-bit multiplier selection signal addressed by a set of these values.
[0105]
The multiplier table 1902 stores eight multiplier values. One of the eight multiplier values is selected by the multiplier selection signal and is output as a multiplier signal.
[0106]
Nine coefficients are stored in the coefficient table 1903, and these values are always output as nine coefficient signals, which are multiplied by multiplier signals by nine multipliers in the multiplier group 1904, respectively. The signal is output as a signal and supplied to the second product-sum operation unit 1605.
[0107]
In this modification, a coefficient signal is generated from the multiplier signal selected according to the identification signal, and thereby the characteristics of the filter can be controlled. Since control is performed using a multiplier signal having one degree of freedom, the degree of freedom of the frequency characteristics of the filter is reduced, but the total size of tables such as a coefficient table and a multiplier table can be reduced.
[0108]
In addition, since the addition is performed after nonlinear conversion or the like after the second product-sum operation process, the high frequency component can be processed independently of the low frequency component.
[0109]
Note that, as an application example of the present invention, the example of switching the filter processing coefficients has been mainly described, but the present invention is not limited to this. For example, it is possible to switch the processing parameters of the color conversion processing and the inking coefficient of the inking processing unit according to the identification signal. This makes it possible to switch the recording chromaticity between color emphasis and density emphasis according to the image area.
[0110]
As described above, according to the embodiment of the present invention, it is possible to finely control the characteristics of image processing such as filter processing according to the area of the image, fine reproduction of fine characters, gradation of silver salt photographs, etc. It is possible to simultaneously achieve sharp reproduction of images and prevention of moire generation at halftone dots.
[0111]
【The invention's effect】
As described above in detail, according to the present invention, the degree of freedom of the filtering process is increased, and the frequency characteristics of the vertical and horizontal images at the time of enlargement / reduction are increased, such as many kinds of regions such as fine halftone dots and coarse halftone dots, fine letters and thick letters. An image processing apparatus that can freely modulate the frequency characteristics of signal processing according to the difference can be provided.
[Brief description of the drawings]
FIG. 1 is a block diagram showing the configuration of a digital color copying machine according to a first embodiment to which an image processing apparatus of the present invention is applied.
FIG. 2 is a diagram illustrating a configuration example of a macro image area identification unit.
FIG. 3 is a diagram for explaining a relationship between a region separation signal value and a region;
FIG. 4 is a diagram illustrating a configuration of a high image quality processing unit.
FIG. 5 is a diagram showing a configuration of a first filter unit.
FIG. 6 is a diagram illustrating a configuration of a delay circuit unit.
FIG. 7 is a diagram illustrating a configuration of a coefficient generation unit.
FIG. 8 is a diagram showing the contents of a table of a coefficient generation unit.
FIG. 9 is a view showing a filter kernel in each region.
FIG. 10 is a diagram showing frequency response characteristics of each region.
FIG. 11 is a diagram for explaining conventional image quality enhancement processing.
FIG. 12 is a diagram showing a conventional filter kernel.
FIG. 13 is a diagram showing frequency characteristics of a conventional filter.
FIG. 14 is a diagram illustrating a filter kernel at 200% enlargement.
FIG. 15 is a diagram illustrating a configuration of a filter unit according to a first modification of the first embodiment.
FIG. 16 is a view showing the contents of a table of a coefficient generation unit.
FIG. 17 is a diagram for explaining an identification signal from a coefficient generation unit.
FIG. 18 is a block diagram showing the configuration of a digital color copying machine according to a second embodiment to which the image processing apparatus of the present invention is applied.
FIG. 19 is a diagram for explaining a relationship between a value of a macro identification signal and a region.
FIG. 20 is a diagram showing a configuration of a micro image area identification unit.
FIG. 21 is a diagram illustrating a configuration of an image quality improvement processing unit.
FIG. 22 is a diagram showing coefficients of a first coefficient generation unit.
FIG. 23 is a diagram showing a configuration of a second coefficient generation unit.
FIG. 24 is a diagram showing a relationship between input and output signals of a lookup table of an identification conversion unit.
FIG. 25 is a diagram for explaining the contents of a coefficient table.
FIG. 26 is a diagram showing characteristics of nonlinear conversion of the adding unit.
FIG. 27 is a diagram illustrating a configuration of a second coefficient generation unit according to a modification.
FIG. 28 is a diagram showing a filter kernel selected in the case of coarse halftone dots and fine halftone dots.
[Explanation of symbols]
101, 1401 ... Image input unit (image input means)
102, 1402 ... color conversion unit
103, 1403 ... Macro image area identification unit (identification means)
104, 1404... Micro image area identification unit (identification means, output means)
105, 1405 ... High image quality processing unit
106, 1406 ... Enlarging / reducing portion
107, 1407 ... Inking process section
108, 1408 ... gradation processing section
109, 1409 ... Image recording section
110, 1410 ... control unit
301, 302, 303 ... Filter section (calculation means)
304 ... Coefficient generator (generating means)
1501, 1502, 1503 ... Image area identification unit
1601 ... Delay circuit section
1602 ... Adder group
1603... First product-sum operation unit (first operation means)
1604: First coefficient generator
1605 ... Second product-sum operation unit (second operation means)
1606: Second coefficient generator (generating means)
1608 ... Adder (addition means)

Claims (3)

First calculation means for separating an input image signal based on density difference and saturation of neighboring pixels and roughness of a halftone dot area;
Extracting structure information that is a density distribution and density change from the input image signal, and referring to the image data separated by the first computing means, a character area, a dot area, and the like for each color component A second computing means for separating into regions;
The first based on the separation signal output from the separated signal and said second computing means is output from the operation unit, the 1-pixel modulation process or the odd-th pixel to record all the pixels in the previous Asked asked before the even A third calculation means for selecting a two-pixel modulation process in which the pixel is recorded in a back-aligned manner;
Have
The third arithmetic means includes a coefficient of a predetermined fixed, a first product-sum operation unit for averaging using degenerate image signal and coefficients of the fixed, pre-Symbol first calculating means and the A second product-sum operation unit that performs a product-sum operation using the second coefficient obtained based on the respective outputs of the second operation means and the degenerate image signal , and the first and second product-sum operations An image processing apparatus comprising: an addition unit that adds and outputs the results calculated by the calculation unit .   The image processing apparatus according to claim 1, wherein the first calculation unit identifies a fine halftone dot region and a coarse halftone dot region using a predetermined frequency as a boundary condition. The image processing apparatus according to claim 2, wherein the third calculation unit changes a filter size based on a calculation result obtained by the first calculation unit .

Images