JP3564789B2 - Color image limited colorizer - Google Patents

Description

色空間変換手段２７から出力されたＬ，ａ，ｂ各成分１画素８ビットの画像データは、それぞれ加算器２１Ｌ，２１ａ，２１ｂに入力され、後述する乗算器２５Ｌ，２５ａ，２５ｂからのデータと加算されて出力される。
【００４１】
加算器２１Ｌ，２１ａ，２１ｂから出力されたＬ，ａ，ｂ各成分の画像データは、それぞれ仮代表色選定手段１２Ｌ，１２ａ，１２ｂに入力され、仮代表色選定手段１２Ｌ，１２ａ，１２ｂにより仮代表色が選定され、それぞれ４ビットのＬ，ａ，ｂ成分の画像データにより表される仮代表色値として出力される。仮代表色選定手段１２Ｌ，１２ａ，１２ｂの機能については、第１の実施例と同様なので、ここでの説明は省略する。この仮代表色選定手段１２Ｌ，１２ａ，１２ｂからのＬ，ａ，ｂ成分の仮代表色値を示すそれぞれ４ビットの画像データは、真代表色選定手段１３に入力される。
【００４２】
この真代表色選定手段１３の機能についても第１の実施例と同様であるため、その説明については省略するが、その入出力特性の一例、すなわちＬ，ａ，ｂ成分の仮代表色値と真代表色番号の関係の一例を図１３（Ａ）に示す。この真代表色選定手段１３からの真代表色番号を示す８ビットのデータは、画像メモリ１４にいったん記憶された後、画像メモリ１４から読み出されてカラーマップ１５に入力される。このカラーマップ１５の機能についても第１の実施例と同様であるため、その説明については省略するが、その一例、すなわち真代表色番号と真代表色の画素値との関係の一例を図１３（Ｂ）に示す。
【００４３】
ここで、仮代表色と真代表色の対応関係を、図５に示したように、仮代表色の中から真代表色を設定する場合について示すと、同図の点Ｐ１で示す加算器２１Ｌ，２１ａ，２１ｂからの出力画素値に対しては、これに対して色差が最も小さい点Ｐ２で示す仮代表色が割り当てられ、その仮代表色に対しては、これに対して色差が最も小さい点Ｐ３，Ｐ４で示す２つの真代表色のうちのたとえば点Ｐ３で示す真代表色が割り当てられる。また、点Ｐ５で示す加算器２１Ｌ，２１ａ，２１ｂからの出力画素値に対しては、これに対して色差が最も小さい点Ｐ６で示す仮代表色が割り当てられ、その仮代表色に対しては、これに対して色差が最も小さい点Ｐ７で示す真代表色が割り当てられる。
【００４４】
具体的に、原画素値と仮代表色の対応関係が図２および図３に示したように設定され、かつ仮代表色と真代表色の対応関係が図１３に示したように設定される場合において、例えば加算器２１Ｌ，２１ａ，２１ｂからの出力画素値がＬ＝１２９，ａ＝９６，ｂ＝１４７であるときには、図１４に示すように、仮代表色選定手段１２Ｌ，１２ａ，１２ｂからの仮代表色はＬ＝８，ａ＝６，ｂ＝９の値のものとなり、その値が真代表色選定手段１３に入力され、真代表色選定手段１３からは真代表色番号「７１」が出力される。真代表色選定手段１３からの出力値である真代表色番号「７１」は画像メモリ１４に入力され、いったん画像メモリ１４に記憶される。画像メモリ１４から読み出された真代表色番号「７１」はカラーマップ１５に入力され、カラーマップ１５から出力される真代表色はＲ＝１１３，Ｇ＝１２８，Ｂ＝６５の値のものとなる。
【００４５】
図１２の第４の実施例においては、カラーマップ１５から読み出されたそれぞれ８ビットのＲ，Ｇ，Ｂ成分の画像データがＤ／Ａ変換器１７Ｒ，１７Ｇ，１７Ｂによってそれぞれアナログ信号に変換され、そのＲ，Ｇ，Ｂ成分の画像信号がＣＲＴディスプレイ１８に供給されて、ＣＲＴディスプレイ１８上に２５６色に限定色化されたカラー画像が表示される。一方、仮代表色選定手段１２Ｌ，１２ａ，１２ｂから出力される仮代表色のＬ，ａ，ｂ成分の画像データは、真代表色選定手段１３に入力されると同時に、ビット数変換器２２Ｌ，２２ａ，２２ｂにも入力される。このビット数変換器２２Ｌ，２２ａ，２２ｂでは、入力値である仮代表色のＬ，ａ，ｂ各成分４ビットの画像データが、各成分８ビットのデータに変換される。このビット数変換器２２Ｌ，２２ａ，２２ｂからのＬ，ａ，ｂ成分の出力値は、差分検出器２３Ｌ，２３ａ，２３ｂにそれぞれ入力される。
【００４６】
Ｌ成分の系についてのみ説明すれば、差分検出器２３Ｌにおいて、ビット変換器２２ＬからのＬ成分のデータと、加算器２１Ｌからのデータとの差分が検出され、リミッタ２４Ｌにおいてその差分データの値がある範囲内に制限され、乗算器２５Ｌにおいてその制限された値の差分データに適当な係数ｋが乗じられ、加算器２１Ｌに入力される。ａ成分の系およびｂ成分の系についても同様で、参照符号中においてＬに代えてａまたはｂを用いて対応部分を示す。このように、最終的な出力値であるカラーマップ１５からの出力画像データと、仮代表色選定手段１２Ｌ，１２ａ，１２ｂから出力される仮代表色のＬ，ａ，ｂ成分の画像データとの差分がフィードバックされることにより、画像全体の平均の色が保存される。
【００４７】
図１５を例にあげてａｂ平面において説明すると、直前の画素についての点Ｐ１１で示す画素値とこれに対して割り当てられた点Ｐ１４で示す真代表色との差αに対して係数ｋが乗じられたものが符号を変えて次の画素についての点Ｐ１２で示す画素値に対して加算され、次の画素については等価的に画素値が点Ｐ１５で示す値とされることにより、次の画素については、例えば点Ｐ１７で示す真代表色が割り当てられることになる。したがって、ある程度以上の色差がある複数の原画素値に対して、集中的に同一の真代表色が割り当てられることがない。
【００４８】
ただし、リミッタ２４Ｌ，２４ａ，２４ｂにおいては上記の差αがあるしきい値以下またはあるしきい値以上に制限され、これにより、フィードバック後の原画素値が本来の原画素値に対して大きくかけ離れてしまって全く異なった色の真代表色が割り当てられてしまうことが防止される。また、上記の差αがあるしきい値以下であるときにはリミッタ２４Ｌ，２４ａ，２４ｂの出力値が０となるようにしてもよく、これによれば下地除去や不要なフィードバックの除去という効果を生じる。また、図示していないが、色空間変換手段２７からの画像データの経路にラインメモリにより近接する複数の画素からなる画素ブロックを形成する回路を設けて、乗算器２５Ｌ，２５ａ，２５ｂからのデータを複数の周辺画素についての画像データに係数ｋを適宜異ならせて加算すると、画質的により効果的である。
【００４９】
上述した第４の実施例によれば、ＲＧＢやＹＭＣなどの非均等色空間に代表されるような限定色化処理および誤差拡散処理を行うに適していない色空間であっても、色空間変換手段２７により、原カラー画像が表現された色空間を、限定色化処理および誤差拡散処理を行うに適した色空間に変換することができるため、誤差拡散における色の保存を視覚的にも満足させた高品質な限定色化を行うことができる。
【００５０】
図１６は、この発明によるカラー画像限定色化装置の第５の実施例を示すブロック図である。この第５の実施例では、第４の実施例において非均等色空間であるＲＧＢ色空間から均等色空間であるＬａｂ色空間に変換した後、Ｌａｂ均等色空間内で差分のフィードバック（誤差拡散）を行っている代わりに、ＲＧＢ色空間内で差分のフィードバックを行う構成となっている。すなわち、画像メモリ１１Ｒ，１１Ｇ，１１Ｂに格納されたＲ，Ｇ，Ｂ成分の画像データは、画像メモリ１１Ｒ，１１Ｇ，１１Ｂから読み出されて画素値変換手段２８に入力される。この画素値変換手段２８は、後述するように、発生した差分を人間の視覚的に正しくフィードバック（誤差拡散）されるように、入力されるＲ，Ｇ，Ｂ成分の画像データを変化させるものである。なお、この画素値変換手段２８の説明については後述する。
【００５１】
画素値変換手段２８から出力されたＲ，Ｇ，Ｂ成分の画像データは、それぞれ加算器２１Ｒ，２１Ｇ，２１Ｂに入力される。後述する画素値変換手段２９から出力されたＲ，Ｇ，Ｂ成分の画像データと加算器２１Ｒ，２１Ｇ，２１Ｂにて加算されて出力される。加算器２１Ｒ，２１Ｇ，２１Ｂから出力されたＲ，Ｇ，Ｂ成分各８ビットの画像データは、それぞれ仮代表色選定手段１２Ｒ，１２Ｇ，１２Ｂに入力され、この仮代表色選定手段１２Ｒ，１２Ｇ，１２Ｂにより仮代表色が選定され、それぞれ４ビットのＲ，Ｇ，Ｂ成分の画像データにより表される仮代表色値として出力される。仮代表色選定手段１２Ｒ，１２Ｇ，１２Ｂの機能については第１の実施例と同様なので、ここでは説明は省略する。
【００５２】
仮代表色選定手段１２Ｒ，１２Ｇ，１２ＢからのＲ，Ｇ，Ｂ成分の仮代表色を示すそれぞれ４ビットの画像データは、真代表色選定手段１３に入力される。この真代表色選定手段１３の機能についても第１の実施例と同様であるため、その説明については省略するが、その入出力特性の一例、すなわちＲ，Ｇ，Ｂ成分の仮代表色値と真代表色値の関係の一例を図１７（Ａ）に示す。この真代表色選定手段１３からの真代表色番号を示す８ビットのデータは、画像メモリ１４にいったん記憶された後、画像メモリ１４から読み出されてカラーマップ１５に入力される。このカラーマップ１５の機能についても第１の実施例と同様であるため、その説明については省略するが、その一例、すなわち真代表色番号と真代表色の画素値との関係の一例を図１７（Ｂ）に示す。
【００５３】
カラーマップ１５から読み出されたそれぞれ８ビットのＲ，Ｇ，Ｂ成分の画像データは、Ｄ／Ａ変換器１７Ｒ，１７Ｇ，１７Ｂによってそれぞれアナログ信号に変換され、そのＲ，Ｇ，Ｂ成分の画像信号がＣＲＴディスプレイ１８に供給されて、ＣＲＴディスプレイ１８上に２５６色に限定色化されたカラー画像が表示される。一方、仮代表色選定手段１２Ｒ，１２Ｇ，１２Ｂから出力される仮代表色のＲ，Ｇ，Ｂ成分の画像データは、真代表色選定手段１３に入力されると同時に、ビット数変換器２２Ｒ，２２Ｇ，２２Ｂにも入力される。このビット数変換器２２Ｒ，２２Ｇ，２２Ｂは、第４の実施例におけるビット数変換器２２Ｌ，２２ａ，２２ｂと同様の機能を持ち、その機能については第４の実施例で説明したのでここでは省略する。このビット数変換器２２Ｒ，２２Ｇ，２２Ｂで８ビットのデータに変換されたＲ，Ｇ，Ｂ成分の画像データは、それぞれ差分検出器２３Ｒ，２３Ｇ，２３Ｂに入力される。
【００５４】
Ｒ成分の系についてのみ説明すれば、差分検出器２３Ｒにおいて、ビット変換器２２Ｒから入力されるＲ成分のデータと、加算器２１Ｒからのデータとの差分が検出され、リミッタ２４Ｒにおいてその差分データの値がある範囲内に制限され、乗算器２５Ｒにおいてその制限された値の差分データに適当な係数ｋが乗じられ、画素値変換手段２９に入力される。Ｇ成分の系およびＢ成分の系についても同様で、参照符号中においてＲに代えてＧまたはＢを用いて対応部分を示す。画素値変換手段２９に入力されたＲ，Ｇ，Ｂ成分の差分の画像データは、画素値変換手段２９において、人間の視覚的に正しくフィードバック（誤差拡散）されるように変換される。この画素値変換手段２９の説明については後述する。画素値変換手段２９から出力されたＲ，Ｇ，Ｂ成分の差分の画像データは、それぞれ加算器２１Ｒ，２１Ｇ，２１Ｂに入力され、加算器２１Ｒ，２１Ｇ，２１Ｂにおいて、画素値変換手段２９から出力されたＲ，Ｇ，Ｂ成分の画像データと加算される。
【００５５】
さて、限定色化を行うにあたっては、先述したように、視覚的に平均の色を保持しながら差分データのフィードバック（誤差拡散）を行うのが望ましい。そこで、上述した第５の実施例では、画素値変換手段２８および画素値変換手段２９を設け、ＲＧＢ非均等色空間において、視覚的に平均の色を保存しながら差分データのフィードバック（誤差拡散）を行うために、画素値変換手段２８および画素値変換手段２９によって画像データおよび差分の画像データの画素値変換を行うようにしている。以下、画素値変換手段２８および画素値変換手段２９について、数式（行列式）を用いながら説明する。
【００５６】
今、ＲＧＢ色空間からＬａｂ色空間への変換は、
【数２】

数２の式と近似することができる。同様に、Ｌａｂ色空間からＲＧＢ色空間への変換も、
【数３】

数３の式と近似することができる。
【００５７】
今、図１８に示すように、任意の画素値をＰ２１＝（Ｒ_１ ，Ｇ_１ ，Ｂ_１）、Ｐ２１の限定色化後の画素値をＰ２２＝（Ｒ_２ ，Ｇ_２ ，Ｂ_２）、Ｐ２１とＰ２２との差分（誤差）が加算される画素の画素値をＰ２３＝（Ｒ_３ ，Ｇ_３ ，Ｂ_３）、Ｐ２３に上記の差分（誤差）が加算された後の画素値をＰ２４＝（Ｒ_４ ，Ｇ_４ ，Ｂ_４）とする。また、ＲＧＢ色空間上で表された上記の画素値Ｐ２１，Ｐ２２，Ｐ２３，Ｐ２４を、Ｌａｂ色空間に変換したときの画素値をＰ２１，Ｐ２２，Ｐ２３，Ｐ２４の順に、（Ｌ_１ ，ａ_１ ，ｂ_１），（Ｌ_２ ，ａ_２ ，ｂ_２），（Ｌ_３ ，ａ_３ ，ｂ_３），（Ｌ_４ ，ａ_４ ，ｂ_４）とする。
【００５８】
すなわち、
【数４】

数４の関係式が成り立つとする。なお、理解を容易にするために、差分（誤差）のフィードバックは隣接する１画素にのみ行うものとし、また乗算器２５Ｒ，２５Ｇ，２５Ｂの乗算係数は全てｋ＝１．０、リミッタ２４Ｒ，２４Ｇ，２４Ｂでは値の制限を行わないものとする。
【００５９】
さて、差分（誤差）のフィードバックはＬａｂ均等色空間で行うことが望ましいので、
【数５】

数５の式の計算をするのがよい。したがって、
【数６】

【数７】

数６および数７の各式の関係が成り立つ。
【００６０】
これにより、画素値変換手段２８は、
【数８】

数８の式のマトリクス演算を行うことのできる加算・乗算器群で構成され、画素値変換手段２９は、
【数９】

数９の式のマトリクス演算を行うことのできる加算・乗算器群で構成される。
【００６１】
また、数６の式の４段目の右辺第１項は、
【数１０】

数１０の式となるため画素値変換手段２８はなくてもよいが、近似誤差をできるだけ少なくするためには画素値変換手段２８を設けることが望ましい。さらに、近似誤差を少なくするためには、画素値変換手段２８および画素値変換手段２９にはＬＵＴを備えることもより有効である。
【００６２】
このように、第５の実施例によれば、仮代表色選定手段１２Ｒ，１２Ｇ，１２Ｂの入力値と出力値との差分を変化させてフィードバック（誤差拡散）させる画素値変換手段２９を設けたことにより、ＲＧＢ非均等色空間であっても視覚的に平均の色を保存して差分のフィードバックを行うことが可能となる。特に、原カラー画像の画素値を変化させて仮代表色選定手段１２Ｒ，１２Ｇ，１２Ｂに与える画素値変換手段２８をも設けることにより、近似誤差をできるだけ少なくすることができるため、限定化処理および誤差拡散処理を行うに適さない色空間においても視覚的に色を保存した誤差拡散を行うことができるため、高品質な限定色化画像を得ることができる。
【００６３】
なお、上記各実施例において、画像メモリ１１Ｌ，１１ａ，１１ｂ（１１Ｒ，１１Ｇ，１１Ｂ）に格納される原カラー画像が１画素１成分につき８ビットで、すなわち１画素あたり合計２４ビットで表現されたものであれば、原カラー画像は約１６７７万色を表すことができ、自然画像を人間の目には全く変わらず表現することができるものの、原カラー画像は必ずしも１画素あたり合計２４ビットで表現されたものである必要はない。仮代表色の総数についても、４０９６色に限る必要はなく、また仮代表色選定手段１２Ｌ，１２ａ，１２ｂ（１２Ｒ，１２Ｇ，１２Ｂ）の入出力特性も図２または図７に示したものとは異なるものにすることができる。
【００６４】
例えば、原カラー画像が１画素１成分につき８ビットで表現されたものである場合に、上述したように、仮代表色選定手段１２Ｌ，１２ａ，１２ｂ（１２Ｒ，１２Ｇ，１２Ｂ）の出力を各成分につき４ビットとするときは、それぞれの出力値として０から１５までの値を得ることができるが、仮代表色選定手段１２Ｌ，１２ａ，１２ｂ（１２Ｒ，１２Ｇ，１２Ｂ）の入出力特性を、入力値範囲は図１９に示すように０付近および２５５付近を除いて均等の幅で、または図２０に示すように中央付近では少ない値の幅で密に、両端側では多い値の幅で疎に、それぞれ合計９個の領域に分割され、この９個の入力値領域に対して出力値は０から８までの値をとる特性とすることもできる。この場合には、仮代表色の総数は９×９×９＝７２９色となり、真代表色選定手段１３の回路規模を小さくすることができる。
【００６５】
このように、仮代表色選定手段１２Ｌ，１２ａ，１２ｂ（１２Ｒ，１２Ｇ，１２Ｂ）を図１９または図２０に示した入出力特性とする場合には、図８の第２の実施例、図１２の第４の実施例および図１６の第５の実施例においては、ビット数変換器２２Ｌ，２２ａ，２２ｂ（２２Ｒ，２２Ｇ，２２Ｂ）は、上述したように、仮代表色選定手段１２Ｌ，１２ａ，１２ｂ（１２Ｒ，１２Ｇ，１２Ｂ）の入力値領域幅の中間値を出力するように、仮代表色選定手段１２Ｌ，１２ａ，１２ｂ（１２Ｒ，１２Ｇ，１２Ｂ）が図１９に示した入出力特性とされるときには図２１（Ａ）に示すような入出力特性のものとし、仮代表色選定手段１２Ｌ，１２ａ，１２ｂ（１２Ｒ，１２Ｇ，１２Ｂ）が図２０に示した入出力特性とされるときには図２１（Ｂ）に示すような入出力特性のものとするのが望ましい。
【００６６】
また、仮代表色選定手段１２Ｌ，１２ａ，１２ｂ（１２Ｒ，１２Ｇ，１２Ｂ）の入出力特性は、仮代表色選定手段１２Ｌ，１２ａ，１２ｂ（１２Ｒ，１２Ｇ，１２Ｂ）ごとに、すなわち各成分ごとに、図２、図６、図１９または図２０に示したような特性の間で互いに異ならせてもよい。さらに、たとえば仮代表色選定手段１２Ｌ，１２ａ，１２ｂ（１２Ｒ，１２Ｇ，１２Ｂ）の出力を各成分につき５ビットとして、仮代表色の総数を３２×３２×３２＝３２７６８色とすることもできる。真代表色の総数も、２５６色に限る必要はない。たとえば、真代表色選定手段１３の出力の真代表色番号を７ビットとして真代表色の総数を１２８色とし、または真代表色番号を９ビットとして真代表色の総数を５１２色とすることができる。
【００６７】
仮代表色選定手段１２Ｌ，１２ａ，１２ｂ（１２Ｒ，１２Ｇ，１２Ｂ）、真代表色選定手段１３、カラーマップ１５および図１１の第３の実施例におけるカラーマップ２６などは、ＬＵＴ以外の手段を用いて構成することができる。たとえば、仮代表色選定手段１２Ｌ，１２ａ，１２ｂ（１２Ｒ，１２Ｇ，１２Ｂ）は、比較器などによって原画素値が分割されたセル空間のいずれに属するかが検出され、その属するセル空間に応じた仮代表色値が出力されるように構成することができる。
【００６８】
【発明の効果】
以上説明したように、この発明によれば、代表色の決定にあたって、あらかじめ原カラー画像が表現される色空間を適宜分割したＮｂ個のセル空間のそれぞれ例えば中心を仮代表色として仮代表色選定手段に登録するとともに、あらかじめＮｃ色の真代表色を設定して真代表色選定手段に登録するので、ヒストグラムの作成を必要とすることなく、少ないメモリ容量と短い処理時間により代表色を決定することができ、また代表色の選定にあたっては、原画素値が上記の分割されたＮｂ個のセル空間のいずれに属するかによって仮代表色選定手段からあらかじめ登録された仮代表色が選定され、次にその仮代表色により真代表色選定手段からあらかじめ登録された真代表色が選定されるので、原画素値と代表色値との間の距離を計算する必要がなく、少ないメモリ容量と短い処理時間により代表色を選定することができる。
【図面の簡単な説明】
【図１】この発明によるカラー画像限定色化装置の第１の実施例を示すブロック図である。
【図２】仮代表色選定手段の入出力特性の一例を示す特性図である。
【図３】その入出力特性をａｂ平面上で示す図である。
【図４】第１の実施例に係る真代表色選定手段およびカラーマップの各入出力特性の一例を示す図である。
【図５】原画素値、仮代表色および真代表色の関係の一例をａｂ平面上で示す図である。
【図６】第１の実施例に係る原画素値、仮代表色および真代表色の関係の具体的数値例を示す図である。
【図７】仮代表色選定手段の入出力特性の他の例を示す特性図である。
【図８】この発明によるカラー画像限定色化装置の第２の実施例を示すブロック図である。
【図９】ビット数変換器の入出力特性の一例を示す図である。
【図１０】第２の実施例のフィードバックの説明に供する図である。
【図１１】この発明によるカラー画像限定色化装置の第３の実施例を示すブロック図である。
【図１２】この発明によるカラー画像限定色化装置の第４の実施例を示すブロック図である。
【図１３】第４の実施例に係る真代表色選定手段およびカラーマップの各入出力特性の一例を示す図である。
【図１４】第４の実施例に係る原画素値、仮代表色および真代表色の関係の具体的数値例を示す図である。
【図１５】第４の実施例のフィードバックの説明に供する図である。
【図１６】この発明によるカラー画像限定色化装置の第５の実施例を示すブロック図である。
【図１７】第５の実施例に係る真代表色選定手段およびカラーマップの各入出力特性の一例を示す図である。
【図１８】第５の実施例のフィードバックの説明に供する図である。
【図１９】仮代表色選定手段の入出力特性のさらに他の例を示す特性図である。
【図２０】仮代表色選定手段の入出力特性のさらに他の例を示す特性図である。
【図２１】ビット数変換器の入出力特性の他の例を示す図である。
【図２２】フルカラー表示のカラー画像表示装置の一例を示すブロック図である。
【図２３】限定色表示のカラー画像表示装置の一例を示すブロック図である。
【図２４】局所区分探索法の説明に供する図である。
【符号の説明】
１２Ｌ，１２ａ，１２ｂ（１２Ｒ，１２Ｇ，１２Ｂ） 仮代表色選定手段
１３ 真代表色選定手段
１５ カラーマップ
２７ 色空間変換手段
２８，２９ 画素値変換手段[0001]
[Industrial applications]
The present invention relates to a color image limited color conversion apparatus, and more particularly to a color image display apparatus or a color image output apparatus capable of simultaneously expressing a digitized full-color image with only a limited number of colors, compared to the original image. The present invention relates to a device for performing limited color so as to express images without any color.
[0002]
[Prior art]
A color image is generally represented by three components of red, green, and blue (so-called three primary colors of light). In order to display a natural image in the same manner as the original image, a digital value of about 8 bits is required as a pixel value of each component. The image memory requires a capacity of about 24 bits per pixel. FIG. 22 shows an example of such a full-color display color image display device. The image memories 1R, 1G, and 1B each have 8 bits of R (red), G (green), and B (blue) for each pixel. Image data is stored, and the R, G, B image data is read out from the image memories 1R, 1G, 1B and converted into analog signals by the D / A converters 2R, 2G, 2B, and the R, G, B images are stored. The image signal is supplied to the CRT display 3, and a full-color image is displayed on the CRT display 3.
[0003]
On the other hand, a personal computer or the like often uses a color image display device that can simultaneously display only a limited number of colors. FIG. 23 shows an example of a color image display device of such a limited color display. The image memory 5 stores color information of 8 bits per pixel, that is, a value from 0 to 255, and stores the color information. The information is read from the image memory 5 and supplied to the color map 6 as its address signal, so that 8-bit R, G, and B image data are read from the color map 6, respectively. Are converted into 8-bit R, G, and B image data, respectively, and the R, G, and B image data are converted into analog signals by D / A converters 7R, 7G, and 7B. The R, G, and B image signals are supplied to the CRT display 8, and a color image of a limited number of 256 colors is displayed on the CRT display 8. Although the color image display device of this example can display only 256 colors at the same time, the capacity of the image memory is only 1/3 that of the color image display device of full color display of the example of FIG.
[0004]
Expressing a full-color image with a limited number of colors is referred to as limiting the color image. However, in general, a natural image cannot be expressed as good as the original image with about 256 colors. However, there is a great demand for making this possible, and various proposals have been made to make this possible. A representative example is the Median Cut Algorithm shown in Computer Graphics, Vol. 16, No. 3, July 1982, pp. 297-307. This is done by examining the color distribution of the original color image in the color space, creating a histogram, and first dividing the color space by the median (median) of the longest axis of the color distribution to form two subspaces (clusters). Then, the two subspaces are divided into four subspaces by dividing each of the two subspaces by the median having the longest axis of color distribution, and a predetermined limited number of color spaces, for example, 256 subspaces are obtained. In each subspace, the average value of the pixel values belonging to that subspace is used as the representative color value, and in limiting the color, the original pixel value is replaced with the representative color value of the final subspace to which it belongs. is there.
[0005]
On the other hand, when the representative color is determined in advance, or when the representative color is determined by some method, it is necessary to replace the original pixel value with the closest representative color value. However, simply calculating the distance between the original pixel value and all the representative color values takes too much time to select the representative color. Therefore, a method of shortening the time required for selecting a representative color has been proposed, and a representative example thereof is a locally sorted search method disclosed in the above-mentioned literature.
[0006]
This means that when the number of representative colors is N, the color space in which the original color image is expressed is divided into N × N × N cell spaces, and a list of representative colors close to the cell space is obtained for each cell space. At this time, among the representative colors indicated by black circles in FIG. 24 (in FIG. 24, the color space is indicated by a two-dimensional plane for convenience), the representative color closest to the center C1 of a certain cell space S1 is defined. In other words, the distance r between the representative color Pa and the corner of the cell space S1 farthest from the representative color Pa in the case of FIG. , The representative colors within the cell space S1 and within a distance r from the cell space S1 are added to the representative color list of the cell space S1, but the representative colors farther than the distance r from the cell space S1 are not added to the representative color list of the cell space S1. To do.
[0007]
That is, in the case of FIG. 24, the representative color list of the cell space S1 includes only the representative colors Pa, Pb, Pc, Pd, Pe, Pf, and Pg located inside the periphery E1 separated by the distance r from the cell space S1. In addition, the representative colors Ph, Pi, etc. located outside the periphery E1 are not added. By doing so, useless representative colors can be excluded from the list. In limiting colorization, first, which cell space the original pixel value belongs to is checked, and then the cell to which the original pixel value belongs is determined. The color closest to the original pixel value is selected from the representative colors in the representative color list of the space.
[0008]
[Problems to be solved by the invention]
However, since the above-described median cut algorithm is a so-called adaptive method of determining a representative color according to the color distribution of an original color image, a high-quality image can be obtained as a color image after limited colorization. To create a histogram by examining the color distribution of the original color image, for example, if the original color image is represented by R, G, B image data of 8 bits per pixel, at least 2²⁴X2 bytes = A memory of about 32 megabytes is required, and furthermore, there is a disadvantage that a computer process requiring a considerable processing time is required to finally determine a representative color by repeating division into partial spaces. .
[0009]
In addition, although the local segment search method requires less time for selecting a representative color than the method of calculating the distance between the original pixel value and all the representative color values, it is considerably more difficult to create a representative color list as preprocessing. And the distance between the original pixel value and a plurality of representative color values in the representative color list must be calculated.
[0010]
Therefore, the present invention is required for determining and selecting a representative color in a color image display device or a color image output device capable of simultaneously expressing (displaying or outputting) a digitized full-color image with only a limited number of colors. It is an object of the present invention to provide a color image limited color conversion apparatus capable of significantly reducing memory capacity and processing time.
[0011]
[Means for Solving the Problems]
A color image limited colorizing apparatus according to the present invention includes a temporary representative color selecting unit that selects a temporary representative color value from a pixel value of an original color image for each component of a color space, and each of the components selected by the temporary representative color selecting unit. And a true representative color selecting means for selecting a true representative color from the temporary representative color values for each component. Here, the “true representative color” referred to in this specification is a color that is ultimately assigned to an original pixel value when the original color image is limited, and is generally referred to as a “representative color”. "Color" or "limited color". The “temporary representative color” is defined as an intermediate value with respect to the original pixel value in the process of finally assigning the true representative color to the original pixel value when the original color image is limited in color in the present invention. When the theoretical number of colors of the original color image is Na, the total number of provisional representative colors is Nb, and the total number of true representative colors is Nc, the relationship of Na> Nb> Nc is satisfied. is there.The temporary representative color selecting means selects a temporary representative color according to which of the Nb cell spaces the original pixel value belongs to.
[0012]
[Action]
In the color image limited color conversion apparatus having the above-mentioned configuration, each of Nb cell spaces obtained by appropriately dividing the color space in which the original color image is expressed is registered in the temporary representative color selection unit, for example, with the center as the temporary representative color. A true representative color of Nc color is set in advance and registered in the true representative color selecting means. In selecting a representative color, a temporary representative color registered in advance by the temporary representative color selecting means is selected according to which of the Nb cell spaces the original pixel value belongs to, and then the temporary representative color is selected. A true representative color registered in advance is selected from the true representative color selecting means according to the representative color.
[0013]
【Example】
Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. In the first to third embodiments, the original color image is L^*a^*b^*Expressed in a uniform color space,
Since each pixel has been digitized to 8 bits per component, the theoretical number of colors of the original color image is 2²⁴Colors, that is, about 16.770,000 colors, and the total number of temporary representative colors is 2¹²= 4096 colors, and the total number of true representative colors is 2⁸= 256 colors will be described as an example. Also, L^*a^*b^*Each component of uniform color space
Is "L^*"" A^*"B^*Should be written in the drawing
However, they are abbreviated as “L”, “a”, and “b” in the specification.
[0014]
FIG. 1 is a block diagram showing a first embodiment of a color image limited colorizing apparatus according to the present invention. In FIG. 1, the image memories 11L, 11a, and 11b store 8 bits of L and a of 8 bits per pixel, which are obtained by reading a color image on a document by an image scanner or the like, or generated by a computer or the like. , B component image data is stored in advance as pixel values of the original color image. The image data of the L, a, and b components stored in the image memories 11L, 11a, and 11b are read out from the image memories 11L, 11a, and 11b, and input to the temporary representative color selection means 12L, 12a, and 12b. As an example, a look-up table (hereinafter, referred to as an LUT) constituted by a RAM or a ROM is used as the temporary representative color selecting means 12L, 12a, 12b, and this is used to store image data from the image memories 11L, 11a, 11b. Are respectively accessed.
[0015]
FIG. 2 shows an example of the input / output characteristics of the provisional representative color selecting means 12L, 12a, 12b, and the input values of values from 0 to 255 each represented by 8-bit L, a, b component image data. The range is divided equally into a range of 17 values except around 0 and 255, and a range of 9 values of about 1/2 of around 0 and 255 is divided into 16 regions in total. Output values of values from 0 to 15 are assigned to the 16 input value regions as temporary representative color values represented by 4-bit L, a, and b component image data. Therefore, the color space in which the original color image is expressed is 2¹²= 4096 cell spaces, and the center of each of the 4096 cell spaces, that is, the color space in which the original color image is represented as the ab plane for convenience, is expressed by the horizontal axis (a-axis) in FIG. The intersection of the horizontal broken line with the vertical (b-axis) and vertical broken lines is set and registered as a temporary representative color. Although only 16 × 16 = 256 colors are shown for convenience in FIG. 3 for the sake of convenience, 16 × 16 × 16 = 4096 colors are actually set in the Lab uniform color space.
[0016]
In this way, a total of 4096 temporary representative colors are set and registered in the temporary representative color selecting means 12L, 12a, 12b, and the temporary representative color selecting means 12L, 12a, 12b stores the original pixel values from the image memories 11L, 11a, 11b. , The temporary representative color selecting means 12L, 12a, and 12b select a temporary representative color having the smallest color difference from the original pixel value among the 4096 temporary representative colors, and each of them has 4 bits. It is output as a temporary representative color value represented by the image data of the L, a, and b components. For example, when the original pixel values are L = 129, a = 96, and b = 147, as is clear from the characteristics of FIG. 2 and as shown on the ab plane of FIG. The provisional representative colors from 12a and 12b have values of L = 8, a = 6, and b = 9.
[0017]
The 4-bit image data indicating the temporary representative color values of the L, a, and b components from the temporary representative color selecting means 12L, 12a, and 12b is input to the true representative color selecting means 13. As an example, an LUT is also used as the true representative color selecting unit 13 and is accessed by the image data from the temporary representative color selecting units 12L, 12a, and 12b. However, the true representative color selecting means 13 does not store the pixel value of the true representative color itself, but sets the pixel value as described later.⁸= 8-bit data indicating the number of true representative colors of 256 colors is stored.
[0018]
FIG. 4A shows an example of the input / output characteristics of the true representative color selecting means 13, that is, an example of the relationship between the temporary representative color values of the L, a, and b components and the true representative color number. The 8-bit data indicating the true representative color number from the true representative color selection unit 13 is temporarily stored in the image memory 14, read out from the image memory 14, and input to the color map 15. The color map 15 is configured by an LUT.⁸= 2 set at address 256 as described later⁸= 256 true representative colors are stored as pixel values represented by 8-bit image data of L, a, and b components, respectively, and are accessed by 8-bit data indicating the true representative color number from the image memory 14. You.
[0019]
FIG. 4B shows an example of the input / output characteristics of the color map 15, that is, an example of the relationship between the true representative color number and the pixel value of the true representative color. As the true representative color, any color up to 256 colors can be set from the color space in which the original color image is expressed, but it is basically preferable to set from the temporary representative colors. That is, when the true representative color is set from the temporary representative colors, as described above, when the original color image is limited by the two-stage limited color conversion, the original pixel value is finally determined. This is because a representative color having a small color difference can be assigned.
[0020]
FIG. 5 shows the relationship between the temporary representative color and the true representative color in the case where the true representative color is set from the temporary representative colors by using the ab plane for convenience, and shows a horizontal broken line and a vertical broken line. Represents a temporary representative color, of which the circled point represents the true representative color. However, a color that cannot be finally reproduced by means for displaying or outputting a color image has no meaning as a true representative color. For example, when a color printer cannot output a color of L = a = b = 0, it is completely meaningless to make such a color a true representative color. Further, even if the colors that can be displayed or output exist, if there are a plurality of colors that cannot be distinguished by humans as true representative colors, they have little meaning except for one color. Therefore, when the means for displaying or outputting is clear, it is desirable to set a color that can be distinguished by humans within the gamut (reproducible color range) of the means as the true representative color.
[0021]
Further, regardless of the type of the original color image, R (red), G (green), B (blue), Y (yellow), M (magenta), C (cyan), K (black), W (white) It is desirable to add a pure color, such as, to the true representative color. In particular, it is more effective to add W (white) to the true representative color. This is because, when W (white) is added to the true representative color, white is assigned to the subtle noise and fog of the white portion included in the original color image at the time of limited color conversion, and the effect of background removal is obtained. This is because Also, depending on the type of the original color image, it is also effective to add an arbitrary color suitable for the image to the true representative color.
[0022]
The conversion from the temporary representative color to the true representative color, an example of which is shown in FIG. 4, is basically performed such that a certain temporary representative color is assigned a true representative color having the smallest color difference. The correspondence between the representative color and the true representative color is determined. However, as described above, when a specific color is set as a true representative color and the specific true representative color is to be assigned instead of the true representative color having the smallest color difference, such a correspondence is established. As shown in FIG. 5, when the true representative color is set from the temporary representative colors as shown in FIG. 5, the correspondence between the temporary representative color and the true representative color is as follows. For this purpose, a temporary representative color indicated by a point P2 having the smallest color difference is assigned to the temporary representative color, and two true representative colors indicated by points P3 and P4 having the smallest color difference are assigned to the temporary representative color. For example, a true representative color indicated by a point P3 is assigned.
[0023]
In addition, a temporary representative color indicated by a point P6 having the smallest color difference is assigned to the original pixel value indicated by the point P5, and a color difference having the smallest color difference is assigned to the temporary representative color. The true representative color indicated by the point P7 is assigned. Specifically, the correspondence between the original pixel values and the temporary representative colors is set as shown in FIGS. 2 and 3, and the correspondence between the temporary representative colors and the true representative colors is set as shown in FIG. In this case, for example, when the original pixel values are L = 129, a = 96, and b = 147, as shown in FIG. 6, the temporary representative color from the temporary representative color selecting means 12L, 12a, 12b is L = 8. , A = 6, b = 9, and the values are input to the true representative color selecting means 13, which outputs the true representative color number “71”. The true representative color number “71”, which is an output value from the true representative color selecting unit 13, is input to the image memory 14 and is temporarily stored in the image memory 14. The true representative color number “71” read from the image memory 14 is input to the color map 15, and the true representative colors output from the color map 15 have values of L = 136, a = 102, and b = 153. Become.
[0024]
In the first embodiment shown in FIG. 1, on the other hand, 8-bit L, a, and b component image data read from the color map 15 are converted into 8-bit R, G, The image data is converted into B component image data, and the R, G, B component image data is converted into analog signals by D / A converters 17R, 17G, 17B, respectively, and the R, G, B component image signals are converted. The color image is supplied to the CRT display 18 and a color image limited to 256 colors is displayed on the CRT display 18. On the other hand, the 8-bit L, a, and b component image data read from the color map 15 are converted into 8-bit Y, M, C, and K component image data by the color conversion unit 19, respectively. The image data of the Y, M, C, and K components is supplied to the color printer 20, and a color image limited to 256 colors is printed out from the color printer 20.
[0025]
According to the first embodiment described above, the temporary representative color selecting means 12L, 12a uses the center of each of the 4096 cell spaces obtained by dividing the Lab uniform color space in which the original color image is rendered substantially equally in advance as the temporary representative color. , 12b, the true representative color numbers from 0 to 255 are set in correspondence with the 4096 temporary representative colors, registered in the true representative color selection means 13, and 256 corresponding to the true representative color numbers. By setting the true representative color of the color and registering it in the color map 15, the representative color can be determined with a small memory capacity and a short processing time without the need to create a histogram.
[0026]
In selecting a representative color, a provisional representative color registered in advance from the provisional representative color selecting means 12L, 12a, 12b is selected depending on which of the 4096 divided cell spaces the original pixel value belongs to. Then, the true representative color number registered in advance from the true representative color selecting means 13 is selected by the temporary representative color, so that the representative color is selected with a small memory capacity and a short processing time without the need for distance calculation. Thus, the original color image is limited in color. In addition, by setting the true representative color based on the above-described consideration, a color image can be displayed or output as inferior to the original image.
[0027]
The input / output characteristics of the tentative representative color selecting means 12L, 12a, 12b are set to 0 as shown in FIG. 7 instead of the input value range being substantially equally divided as shown in FIG. To 255, the input value range can be set to a non-linear characteristic which is divided into small values near the center with a small value width and sparsely provided with a large value width at both ends. With such characteristics, a large number of temporary representative colors can be assigned near the center of the color space that is generally included in a natural image.
[0028]
FIG. 8 is a block diagram showing a second embodiment of the color image limited colorizing apparatus according to the present invention. The second embodiment is different from the first embodiment in FIG. 1 in that the provisional representative color value from the provisional representative color selecting means is fed back to the original pixel value. That is, for the L component system, the adder 21L adds the data from the multiplier 25L described later to the 8-bit image data from the image memory 11L, and the bit number converter 22L outputs the data from the temporary representative color selecting unit 12L. The image data is converted from 4 bits to 8 bits, and a difference detector 23L detects a difference between the image data input from the adder 21L to the temporary representative color selecting unit 12L and the image data from the bit number converter 22L, In the limiter 24L, the value of the difference data is limited to a certain range, and in the multiplier 25L, the difference data of the limited value is multiplied by an appropriate coefficient k. The same applies to the system of the component a and the system of the component b. Corresponding parts are indicated by using a or b instead of L in the reference numerals.
[0029]
The bit number converters 22L, 22a, and 22b may be configured to convert the number of bits by a simple bit shift, for example, when the input value is “0110”, the output value is “011000000”. It is desirable to have input / output characteristics opposite to those of the temporary representative color selecting means 12L, 12a, 12b so as to output an intermediate value of the input value area width of the temporary representative color selecting means 12L, 12a, 12b. . That is, when the provisional representative color selecting means 12L, 12a, 12b has the input / output characteristics shown in FIG. 2, the bit number converters 22L, 22a, 22b are provided with the input / output characteristics as shown in FIG. If the provisional representative color selecting means 12L, 12a, 12b have the input / output characteristics shown in FIG. 7, the bit number converters 22L, 22a, 22b are provided with input / output characteristics as shown in FIG. It is preferable to have output characteristics.
[0030]
In the first embodiment of FIG. 1, if the color space in which the original color image is represented is represented as an ab plane for convenience, the original pixel values indicated by points P11 and P12 in FIG. A plurality of original pixel values, such as a temporary representative color indicated by a point P13 which is the temporary representative color having the smallest color difference, and a true representative color indicated by a point P14 are assigned to the temporary representative color, for example. , The same true representative color is intensively assigned to the plurality of original pixel values, so that an image different from the original color image is displayed or output. It is possible.
[0031]
On the other hand, in the second embodiment of FIG. 8, the difference α between the original pixel value of the immediately preceding pixel indicated by the point P11 and the temporary representative color value indicated by the point P13 assigned thereto is calculated. The value multiplied by the coefficient k changes the sign and is added to the original pixel value of the next pixel indicated by the point P12. For the next pixel, the original pixel value is equivalently set to the value indicated by the point P15. Accordingly, for the next pixel, for example, a temporary representative color indicated by a point P16 is assigned, and a true representative color indicated by a point P17 is assigned. Therefore, the same true representative color is not intensively assigned to a plurality of original pixel values having a certain or more color difference.
[0032]
However, in the limiters 24L, 24a, and 24b, the difference α is limited to a certain threshold value or less, or a certain threshold value or more, whereby the original pixel value after feedback greatly differs from the original original pixel value. As a result, it is possible to prevent a completely different true representative color from being assigned. When the difference α is equal to or less than a certain threshold value, the output values of the limiters 24L, 24a, 24b may be set to 0, thereby producing the effect of removing the background and removing unnecessary feedback. . However, depending on the value of the coefficient k multiplied by the multipliers 25L, 25a, 25b, the limiters 24L, 24a, 24b may not be provided. It is desirable that the value of the coefficient k is 0 or more and 1 or less.
[0033]
Although not shown, a circuit for forming a pixel block composed of a plurality of pixels closer to the line memory in the path of the image data from the image memories 11L, 11a, and 11b is provided, and the circuits are provided from the multipliers 25L, 25a, and 25b. Is added to the image data of a plurality of peripheral pixels by appropriately changing the coefficient k, it is more effective in terms of image quality. Although not shown, also in the second embodiment, a color image of only 256 colors can be displayed on the CRT display as in the first embodiment of FIG.
[0034]
FIG. 11 is a block diagram showing a third embodiment of the color image limited colorizing apparatus according to the present invention. In the third embodiment, the true representative color value based on the true representative color number from the true representative color selecting means is fed back to the original pixel value in the configuration of the first embodiment in FIG. . That is, the data indicating the true representative color number from the true representative color selecting unit 13 is input to the color map 26 similar to the color map 15, and the L, a, and b components indicating the pixel values of the true representative color are obtained from the color map 26. Are read out, and in the difference detectors 23L, 23a, and 23b, the image data of the L, a, and b components input from the adders 21L, 21a, and 21b to the tentative representative color selecting means 12L, 12a, and 12b are used. The difference between the image data of the L, a, and b components from the color map 26, that is, the difference between the pixel value before limited colorization and the pixel value after limited colorization is detected. Other configurations are the same as those of the second embodiment in FIG.
[0035]
Also in the third embodiment, if the data from the multipliers 25L, 25a, and 25b are added to the image data for a plurality of peripheral pixels, it is more effective in terms of image quality. In the third embodiment of FIG. 11, the same effects as those of the second embodiment of FIG. 8 are produced, but when both are compared, in the case of the second embodiment of FIG. The color map 26 as in the third embodiment is not required, and the circuit scale can be reduced as compared with the third embodiment in FIG. On the other hand, in the case of the third embodiment shown in FIG. 11, the difference between the original pixel value and the true representative color finally assigned to the original pixel value is fed back to the original pixel value, and the difference between the original pixel value and the limited color The accurate difference is fed back even when the true representative color registered and set in the image is assigned, so that the density of the entire image can be saved, and the image quality is limited as compared with the second embodiment of FIG. A colored image can be obtained.
[0036]
Here, as in the case of the second embodiment, the difference between the temporary representative color value from the temporary representative color selecting means 121, 12a, and 12b and the pixel value of the original color image is fed back to the pixel value of the original color image. Or a method of feeding back the difference between the true representative color value output from the true representative color selecting unit 13 and the pixel value of the original color image to the pixel value of the original color image as in the third embodiment. Is a so-called error diffusion method. That is, in this error diffusion method, the average density of an arbitrary region of the original color image and an arbitrary region of the image after error diffusion are equalized (saved) by distributing an error generated in a certain pixel to a plurality of surrounding pixels, and Is being improved.
[0037]
By the way, as is common to the embodiments of FIGS. 1, 8 and 11, the color space in which the original color image is expressed is assigned a temporary representative color having the smallest color difference with respect to the original pixel value at the time of limiting colorization. In order to assign a true representative color with a small color difference, the Lab uniform color space described above or the Luv uniform color space also recommended by the CIE (International Commission on Illumination), in which the Euclidean distance in the color space is visually equal to the color difference. However, the present invention is not limited to this, and may be a non-uniform color space such as an RBG color space or a YMC color space. However, in the non-uniform color space such as RGB or YMC, the average of the colors is equivalent in numerical data (although the colors are preserved), but the colors are not necessarily preserved visually by humans. This was not considered to be the case and it was not suitable for performing limited color processing and error diffusion.
[0038]
Therefore, fourth and fifth embodiments in which limited color processing and error diffusion processing can be performed even in this non-uniform color space will be described below. In the fourth and fifth embodiments, an RGB non-uniform color space is taken as an example of a color space unsuitable for performing limited color processing and error diffusion processing, and Lab is used as a color space suitable for performing limited color processing and error diffusion processing. The uniform color space is taken as an example. Note that the original color image is expressed in an RGB non-uniform color space and is digitized to 8 bits per one pixel component.²⁴Color, that is, about 16.770,000 colors, and the total number of temporary representative colors is 2¹²= 4096 colors, and the total number of true representative colors is 2⁸= 256 colors will be described as an example.
[0039]
FIG. 12 is a block diagram showing a fourth embodiment of the color image limited colorizing apparatus according to the present invention. In the drawing, the same parts as those in FIGS. 1 and 8 are denoted by the same reference numerals. , And a detailed description thereof will be omitted because it is redundant. In FIG. 12, the image memories 11R, 11G, and 11B have R, G of 8 bits per pixel obtained by reading a color image on a document by an image scanner or the like or generated by a computer or the like. , B component image data are stored in advance as pixel values of the original color image. The image data of the R, G, and B components stored in the image memories 11R, 11G, and 11B are read from the image memories 11R, 11G, and 11B and input to the color space conversion unit 27.
[0040]
The color space conversion means 27 converts the input R, G, B component image data of 8 bits per pixel into image data of 8 × 3 = 24 bits per pixel expressed in the Lab uniform color space. Processing is performed. As the color space conversion means 27, a known technique may be used. As an example, an LUT constituted by a RAM or a ROM and an adder / multiplier group capable of performing a matrix operation of Expression 1 are used. The following configuration can be used.
(Equation 1)

The 8-bit image data of each pixel of each component of L, a, and b output from the color space conversion means 27 is input to adders 21L, 21a, and 21b, respectively, where the data from the multipliers 25L, 25a, and 25b, which will be described later, It is added and output.
[0041]
The image data of each of the L, a, and b components output from the adders 21L, 21a, and 21b are input to the temporary representative color selecting means 12L, 12a, and 12b, respectively, and are temporarily output by the temporary representative color selecting means 12L, 12a, and 12b. A representative color is selected and output as a provisional representative color value represented by 4-bit L, a, and b component image data. The functions of the tentative representative color selecting means 12L, 12a, 12b are the same as those of the first embodiment, and the description is omitted here. The 4-bit image data indicating the temporary representative color values of the L, a, and b components from the temporary representative color selecting means 12L, 12a, and 12b is input to the true representative color selecting means 13.
[0042]
Since the function of the true representative color selecting means 13 is the same as that of the first embodiment, the description thereof is omitted, but an example of its input / output characteristics, that is, the temporary representative color values of the L, a, and b components and FIG. 13A shows an example of the relationship between true representative color numbers. The 8-bit data indicating the true representative color number from the true representative color selection unit 13 is temporarily stored in the image memory 14, read out from the image memory 14, and input to the color map 15. Since the function of the color map 15 is the same as that of the first embodiment, a description thereof will be omitted, but an example thereof, that is, an example of the relationship between the true representative color number and the pixel value of the true representative color is shown in FIG. It is shown in (B).
[0043]
Here, as shown in FIG. 5, the correspondence between the tentative representative color and the true representative color is shown in a case where the true representative color is set from among the tentative representative colors. , 21a, and 21b are assigned a temporary representative color indicated by a point P2 having the smallest color difference, and the temporary representative color has the smallest color difference. Of the two true representative colors indicated by the points P3 and P4, for example, the true representative color indicated by the point P3 is assigned. Further, a temporary representative color indicated by a point P6 having the smallest color difference is assigned to the output pixel value from the adders 21L, 21a, 21b indicated by a point P5. The true representative color indicated by the point P7 having the smallest color difference is assigned to this.
[0044]
Specifically, the correspondence between the original pixel value and the temporary representative color is set as shown in FIGS. 2 and 3, and the correspondence between the temporary representative color and the true representative color is set as shown in FIG. In the case, for example, when the output pixel values from the adders 21L, 21a, 21b are L = 129, a = 96, b = 147, as shown in FIG. 14, the temporary representative color selecting means 12L, 12a, 12b Are representative values of L = 8, a = 6, and b = 9, and the values are input to the true representative color selecting unit 13, and the true representative color selecting unit 13 outputs the true representative color number “71”. Is output. The true representative color number “71”, which is an output value from the true representative color selecting unit 13, is input to the image memory 14 and is temporarily stored in the image memory 14. The true representative color number “71” read from the image memory 14 is input to the color map 15, and the true representative colors output from the color map 15 have values of R = 113, G = 128, and B = 65. Become.
[0045]
In the fourth embodiment shown in FIG. 12, 8-bit R, G, and B component image data read from the color map 15 are converted into analog signals by D / A converters 17R, 17G, and 17B, respectively. The image signals of the R, G, and B components are supplied to the CRT display 18, and a color image limited to 256 colors is displayed on the CRT display 18. On the other hand, the image data of the L, a, and b components of the tentative representative color output from the tentative representative color selecting means 12L, 12a, and 12b is input to the true representative color selecting means 13 and at the same time, the bit number converter 22L, The data is also input to 22a and 22b. In the bit number converters 22L, 22a, and 22b, the image data of the L, a, and b components of the temporary representative color, which are input values, is converted into 8-bit data of each component. The output values of the L, a, and b components from the bit number converters 22L, 22a, and 22b are input to difference detectors 23L, 23a, and 23b, respectively.
[0046]
If only the system of the L component is described, the difference between the data of the L component from the bit converter 22L and the data from the adder 21L is detected by the difference detector 23L, and the value of the difference data is detected by the limiter 24L. The difference data is limited to a certain range, and the difference data of the limited value is multiplied by an appropriate coefficient k in the multiplier 25L and input to the adder 21L. The same applies to the system of the component a and the system of the component b. Corresponding parts are indicated by using a or b instead of L in the reference numerals. As described above, the output image data from the color map 15, which is the final output value, and the image data of the L, a, and b components of the temporary representative colors output from the temporary representative color selecting means 12L, 12a, and 12b. By feeding back the difference, the average color of the entire image is preserved.
[0047]
To explain in the ab plane by taking FIG. 15 as an example, the coefficient k is multiplied by the difference α between the pixel value of the immediately preceding pixel indicated by the point P11 and the true representative color indicated by the point P14 assigned thereto. The sign of the next pixel is changed and added to the pixel value of the next pixel indicated by the point P12, and the pixel value of the next pixel is equivalently set to the value indicated by the point P15. For example, a true representative color indicated by a point P17 is assigned. Therefore, the same true representative color is not intensively assigned to a plurality of original pixel values having a certain or more color difference.
[0048]
However, in the limiters 24L, 24a, and 24b, the difference α is limited to a certain threshold value or less, or a certain threshold value or more, whereby the original pixel value after feedback greatly differs from the original original pixel value. As a result, it is possible to prevent a completely different true representative color from being assigned. When the difference α is equal to or less than a certain threshold value, the output values of the limiters 24L, 24a, 24b may be set to 0, thereby producing the effect of removing the background and removing unnecessary feedback. . Although not shown, a circuit for forming a pixel block composed of a plurality of pixels closer to the line memory on the path of the image data from the color space conversion means 27 is provided, and the data from the multipliers 25L, 25a, and 25b are provided. Is added to the image data of a plurality of peripheral pixels by appropriately changing the coefficient k, which is more effective in terms of image quality.
[0049]
According to the above-described fourth embodiment, even if the color space is not suitable for performing limited colorization processing and error diffusion processing as represented by non-uniform color spaces such as RGB and YMC, color space conversion is performed. By means 27, the color space in which the original color image is expressed can be converted into a color space suitable for performing the limited colorization process and the error diffusion process, so that the color storage in the error diffusion is visually satisfied. It is possible to perform high quality limited colorization.
[0050]
FIG. 16 is a block diagram showing a fifth embodiment of the color image limited colorizing apparatus according to the present invention. In the fifth embodiment, after the RGB color space which is a non-uniform color space is converted to the Lab color space which is a uniform color space in the fourth embodiment, a difference feedback (error diffusion) is performed in the Lab uniform color space. Is performed, the difference is fed back in the RGB color space. That is, the image data of the R, G, and B components stored in the image memories 11R, 11G, and 11B are read from the image memories 11R, 11G, and 11B and input to the pixel value conversion unit 28. As will be described later, the pixel value conversion means 28 changes the input R, G, B component image data so that the generated difference can be visually fed back (error diffusion) to human eyes. is there. The description of the pixel value converter 28 will be described later.
[0051]
The image data of the R, G, and B components output from the pixel value conversion means 28 are input to adders 21R, 21G, and 21B, respectively. The image data of the R, G, and B components output from the pixel value conversion means 29 described later are added and output by the adders 21R, 21G, and 21B. The 8-bit image data of each of the R, G, and B components output from the adders 21R, 21G, and 21B is input to the tentative representative color selecting means 12R, 12G, and 12B, respectively. A temporary representative color is selected by 12B, and is output as a temporary representative color value represented by image data of R, G, and B components of 4 bits each. The functions of the tentative representative color selecting means 12R, 12G, and 12B are the same as those in the first embodiment, and thus description thereof is omitted.
[0052]
The 4-bit image data indicating the temporary representative colors of the R, G, and B components from the temporary representative color selecting units 12R, 12G, and 12B is input to the true representative color selecting unit 13. Since the function of the true representative color selecting means 13 is the same as that of the first embodiment, the description thereof is omitted, but an example of its input / output characteristics, that is, the temporary representative color values of the R, G, B components and FIG. 17A shows an example of the relationship between true representative color values. The 8-bit data indicating the true representative color number from the true representative color selection unit 13 is temporarily stored in the image memory 14, read out from the image memory 14, and input to the color map 15. Since the function of the color map 15 is the same as that of the first embodiment, the description thereof is omitted, but an example thereof, that is, an example of the relationship between the true representative color number and the pixel value of the true representative color is shown in FIG. It is shown in (B).
[0053]
The 8-bit R, G, B component image data read from the color map 15 is converted into analog signals by D / A converters 17R, 17G, 17B, respectively, and the R, G, B component images are converted. The signal is supplied to the CRT display 18, and a color image limited to 256 colors is displayed on the CRT display 18. On the other hand, the image data of the R, G, and B components of the tentative representative color output from the tentative representative color selecting means 12R, 12G, and 12B is input to the true representative color selecting means 13 and at the same time, the bit number converter 22R, 22G and 22B are also input. The bit number converters 22R, 22G, and 22B have the same functions as those of the bit number converters 22L, 22a, and 22b in the fourth embodiment, and the functions thereof have been described in the fourth embodiment, and thus are omitted here. I do. The R, G, and B component image data converted into 8-bit data by the bit number converters 22R, 22G, and 22B are input to difference detectors 23R, 23G, and 23B, respectively.
[0054]
Only the system of the R component will be described. In the difference detector 23R, a difference between the data of the R component input from the bit converter 22R and the data from the adder 21R is detected, and the difference data of the difference data is detected by the limiter 24R. The value is restricted within a certain range, and the difference data of the restricted value is multiplied by an appropriate coefficient k in the multiplier 25R, and is input to the pixel value conversion means 29. The same applies to the G component system and the B component system, and corresponding parts are indicated by using G or B in place of R in the reference numerals. The image data of the difference between the R, G, and B components input to the pixel value conversion unit 29 is converted by the pixel value conversion unit 29 so that human feedback is visually correct (error diffusion). The description of the pixel value conversion means 29 will be described later. The image data of the difference between the R, G, and B components output from the pixel value conversion means 29 is input to adders 21R, 21G, and 21B, respectively, and output from the pixel value conversion means 29 in the adders 21R, 21G, and 21B. Is added to the image data of the R, G, and B components.
[0055]
When performing limited color conversion, as described above, it is desirable to perform feedback (error diffusion) of difference data while visually maintaining the average color. Therefore, in the fifth embodiment described above, the pixel value conversion means 28 and the pixel value conversion means 29 are provided, and the difference data is fed back (error diffusion) while visually maintaining the average color in the RGB non-uniform color space. The pixel value conversion means 28 and the pixel value conversion means 29 perform the pixel value conversion of the image data and the difference image data. Hereinafter, the pixel value conversion means 28 and the pixel value conversion means 29 will be described using mathematical formulas (determinants).
[0056]
Now, the conversion from the RGB color space to the Lab color space is as follows.
(Equation 2)

Equation 2 can be approximated. Similarly, the conversion from the Lab color space to the RGB color space is
(Equation 3)

Equation 3 can be approximated.
[0057]
Now, as shown in FIG. 18, an arbitrary pixel value is set to P21 = (R₁, G₁, B₁), P21 = (R₂, G₂, B₂), The pixel value of the pixel to which the difference (error) between P21 and P22 is added is P23 = (R₃, G₃, B₃), The pixel value after the above difference (error) is added to P23 is P24 = (R₄, G₄, B₄). The pixel values obtained by converting the pixel values P21, P22, P23, and P24 represented on the RGB color space into the Lab color space are represented by (L) in the order of P21, P22, P23, and P24.₁, A₁, B₁), (L₂, A₂, B₂), (L₃, A₃, B₃), (L₄, A₄, B₄).
[0058]
That is,
(Equation 4)

It is assumed that the relational expression of Expression 4 holds. In order to facilitate understanding, feedback of the difference (error) is performed only for one adjacent pixel, and the multiplier coefficients of the multipliers 25R, 25G, and 25B are all k = 1.0, and the limiters 24R and 24G. , 24B do not limit the value.
[0059]
Now, it is desirable that the difference (error) feedback is performed in the Lab uniform color space.
(Equation 5)

It is good to calculate the formula of Formula 5. Therefore,
(Equation 6)

(Equation 7)

The relations of the equations (6) and (7) hold.
[0060]
Thereby, the pixel value conversion means 28
(Equation 8)

The pixel value conversion means 29 is composed of a group of adders / multipliers capable of performing the matrix operation of the equation (8).
(Equation 9)

It is composed of an adder / multiplier group capable of performing the matrix operation of the equation (9).
[0061]
In addition, the first term on the right side of the fourth stage of the equation (6) is
(Equation 10)

Although the pixel value conversion means 28 may be omitted because of the formula (10), it is desirable to provide the pixel value conversion means 28 in order to minimize the approximation error. Further, in order to reduce the approximation error, it is more effective to provide the pixel value converter 28 and the pixel value converter 29 with an LUT.
[0062]
As described above, according to the fifth embodiment, the pixel value conversion unit 29 for changing the difference between the input value and the output value of the provisional representative color selection units 12R, 12G, and 12B and performing feedback (error diffusion) is provided. As a result, even in the RGB non-uniform color space, it is possible to visually save the average color and feed back the difference. In particular, the provision of the pixel value conversion means 28 for changing the pixel value of the original color image and giving it to the provisional representative color selection means 12R, 12G, 12B can reduce the approximation error as much as possible. Even in a color space that is not suitable for performing the error diffusion process, it is possible to perform the error diffusion while visually preserving the color, so that a high-quality limited color image can be obtained.
[0063]
In each of the above embodiments, the original color image stored in the image memories 11L, 11a, and 11b (11R, 11G, and 11B) is represented by 8 bits per one pixel, that is, a total of 24 bits per pixel. , The original color image can represent about 16.77 million colors, and a natural image can be represented without any change to the human eye, but the original color image is always represented by a total of 24 bits per pixel. It doesn't have to be. The total number of temporary representative colors does not need to be limited to 4096 colors, and the input / output characteristics of the temporary representative color selecting means 12L, 12a, 12b (12R, 12G, 12B) are different from those shown in FIG. 2 or FIG. It can be different.
[0064]
For example, when the original color image is represented by 8 bits per pixel, as described above, the outputs of the provisional representative color selecting means 12L, 12a, 12b (12R, 12G, 12B) are used for each component. When each bit is 4 bits, values from 0 to 15 can be obtained as output values. However, the input / output characteristics of the temporary representative color selecting means 12L, 12a, 12b (12R, 12G, 12B) are input. The value range is a uniform width except for around 0 and 255 as shown in FIG. 19, or densely with a small value width near the center and sparsely with a large value width at both ends as shown in FIG. , Each of which is divided into a total of nine regions, and the output values of the nine input value regions can take values from 0 to 8. In this case, the total number of temporary representative colors is 9 × 9 × 9 = 729, and the circuit scale of the true representative color selecting unit 13 can be reduced.
[0065]
As described above, when the temporary representative color selecting means 12L, 12a, 12b (12R, 12G, 12B) has the input / output characteristics shown in FIG. 19 or FIG. 20, the second embodiment of FIG. In the fourth embodiment and the fifth embodiment shown in FIG. 16, the bit number converters 22L, 22a, 22b (22R, 22G, 22B), as described above, use the temporary representative color selecting means 12L, 12a, The provisional representative color selecting means 12L, 12a, 12b (12R, 12G, 12B) have the input / output characteristics shown in FIG. 19 so as to output an intermediate value of the input value area width of 12b (12R, 12G, 12B). When the temporary representative color selecting means 12L, 12a, 12b (12R, 12G, 12B) has the input / output characteristics shown in FIG. 20, the input / output characteristics shown in FIG. (B) It is desirable to those of Suyo input-output characteristics.
[0066]
The input / output characteristics of the temporary representative color selecting means 12L, 12a, 12b (12R, 12G, 12B) are different for each of the temporary representative color selecting means 12L, 12a, 12b (12R, 12G, 12B), that is, for each component. , FIG. 2, FIG. 6, FIG. 19 or FIG. Further, for example, the output of the temporary representative color selecting means 12L, 12a, 12b (12R, 12G, 12B) may be set to 5 bits for each component, and the total number of temporary representative colors may be 32 × 32 × 32 = 32768 colors. The total number of true representative colors need not be limited to 256 colors. For example, the true representative color number output from the true representative color selecting means 13 is set to 7 bits and the total number of true representative colors is set to 128 colors, or the true representative color number is set to 9 bits and the total number of true representative colors is set to 512 colors. it can.
[0067]
The temporary representative color selecting means 12L, 12a, 12b (12R, 12G, 12B), the true representative color selecting means 13, the color map 15, and the color map 26 in the third embodiment of FIG. Can be configured. For example, the provisional representative color selection means 12L, 12a, 12b (12R, 12G, 12B) detects which of the divided cell spaces the original pixel value belongs to by a comparator or the like, and according to the cell space to which the original pixel value belongs. It can be configured so that the temporary representative color value is output.
[0068]
【The invention's effect】
As described above, according to the present invention, when determining a representative color, a temporary representative color is selected as a temporary representative color, for example, at the center of each of Nb cell spaces obtained by appropriately dividing a color space in which an original color image is expressed. Since the true representative colors of Nc colors are set in advance and registered in the true representative color selecting means, the representative colors are determined with a small memory capacity and a short processing time without the need to create a histogram. In selecting a representative color, a temporary representative color registered in advance by a temporary representative color selecting unit is selected according to which of the Nb divided cell spaces the original pixel value belongs to. Since the true representative color registered in advance is selected from the true representative color selecting means based on the temporary representative color, it is necessary to calculate the distance between the original pixel value and the representative color value. No, it is possible to select a representative color by a small memory capacity and a short processing time.
[Brief description of the drawings]
FIG. 1 is a block diagram showing a first embodiment of a color image limited colorizing apparatus according to the present invention.
FIG. 2 is a characteristic diagram illustrating an example of input / output characteristics of a temporary representative color selecting unit.
FIG. 3 is a diagram showing the input / output characteristics on an ab plane.
FIG. 4 is a diagram illustrating an example of each input / output characteristic of a true representative color selecting unit and a color map according to the first embodiment.
FIG. 5 is a diagram illustrating an example of a relationship between an original pixel value, a temporary representative color, and a true representative color on an ab plane.
FIG. 6 is a diagram illustrating a specific numerical example of a relationship between an original pixel value, a temporary representative color, and a true representative color according to the first embodiment.
FIG. 7 is a characteristic diagram showing another example of the input / output characteristics of the temporary representative color selecting unit.
FIG. 8 is a block diagram showing a second embodiment of the color image limited colorizing apparatus according to the present invention.
FIG. 9 is a diagram illustrating an example of input / output characteristics of a bit number converter.
FIG. 10 is a diagram for explaining feedback of the second embodiment;
FIG. 11 is a block diagram showing a third embodiment of the color image limited colorizing apparatus according to the present invention.
FIG. 12 is a block diagram showing a fourth embodiment of the color image limited colorizing apparatus according to the present invention.
FIG. 13 is a diagram illustrating an example of each input / output characteristic of a true representative color selecting unit and a color map according to a fourth embodiment.
FIG. 14 is a diagram showing specific numerical examples of the relationship between an original pixel value, a temporary representative color, and a true representative color according to the fourth embodiment.
FIG. 15 is a diagram provided for explanation of feedback in the fourth embodiment.
FIG. 16 is a block diagram showing a fifth embodiment of the color image limited colorizing apparatus according to the present invention.
FIG. 17 is a diagram illustrating an example of each input / output characteristic of a true representative color selecting unit and a color map according to a fifth embodiment.
FIG. 18 is a diagram provided for explanation of feedback in the fifth embodiment.
FIG. 19 is a characteristic diagram showing still another example of the input / output characteristics of the temporary representative color selecting unit.
FIG. 20 is a characteristic diagram showing still another example of the input / output characteristics of the temporary representative color selecting unit.
FIG. 21 is a diagram illustrating another example of the input / output characteristics of the bit number converter.
FIG. 22 is a block diagram illustrating an example of a color image display device for full-color display.
FIG. 23 is a block diagram showing an example of a color image display device for limited color display.
FIG. 24 is a diagram provided for explanation of a local segment search method.
[Explanation of symbols]
12L, 12a, 12b (12R, 12G, 12B) Temporary representative color selecting means
13 True representative color selection means
15 Color Map
27 Color space conversion means
28, 29 pixel value conversion means

Claims (6)

When the number of colors of the original color image is Na, the total number of temporary representative colors is Nb, and the total number of true representative colors is Nc from the pixel values of the original color image for each component of the color space , Na>Nb> Nc Provisional representative color selecting means for selecting a temporary representative color value to be related according to which of the Nb cell spaces the pixel value belongs to;
A true representative color selecting unit for selecting a true representative color from the temporary representative color values for each component selected by the temporary representative color selecting unit. 2. The color image limited colorizing apparatus according to claim 1, further comprising a feedback unit that feeds back a difference between an input value and an output value of the temporary representative color selection unit to a pixel value of an original color image. 2. A color image according to claim 1, further comprising feedback means for feeding back a difference between an input value of said temporary representative color selection means and an output value of said true representative color selection means to a pixel value of an original color image. Limited colorization device. A color space conversion unit that converts the pixel value of the color space of the original color image into a pixel value of a color space suitable for performing the limited color conversion process and the error diffusion process and provides the pixel value to the temporary representative color selection unit;
3. The color image limited colorization apparatus according to claim 2, wherein said feedback means feeds back a difference between an input value and an output value of said temporary representative color selection means to an output value of said color space conversion means. 3. The color image limited color conversion according to claim 2, wherein said feedback means changes a difference between an input value and an output value of said temporary representative color selection means and feeds back the pixel value of the original color image. apparatus. Pixel value conversion means for changing the pixel value of the original color image, wherein the feedback means changes the difference between the input value and the output value of the provisional representative color selection means to output the pixel value conversion means 6. The color image limited color conversion apparatus according to claim 5, wherein feedback is provided.

Images