JP2004208128A - Data conversion method and data converter - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To reduce the size of a used memory in a data conversion method for converting input color data into output color data using a lookup table. <P>SOLUTION: A plurality of respective lattice point data in an original lookup table are merged based on a predetermined criterion to acquire merged data and the merged data are stored in a memory as lattice point data in a new lookup table. These merged data are read from the memory and separated into original individual lattice point data. Then, these data are subjected to usual data conversion processing. For example, all of Y bits of the lattice point data for four output colors Y, M, C, K are merged to make the bitwidth 4×Y bits. Upon data conversion processing, 4×Y bits of the merged data are read from the memory, and are separated into Y bits of lattice point data for respective output colors Y, M, C, K. <P>COPYRIGHT: (C)2004,JPO&NCIPI

Description

【００５５】
また式（１）を満足する限り、図２（Ｂ）や図２（Ｃ）に示すように、複数のグループに分けて纏めることもできる。また、図２（Ｃ）に示すように、纏め方によっては、１構成要素のデータビット幅そのもので格納されるグループができる。なお、複数のグループに分けるということから、この場合、出力構成要素数Ｘが３以上であることが前提となる。纏めるからには、出力構成要素数Ｘが複数（２以上）でなければならず、ｘ＝２の場合、これを纏めると１つのグループしかできなからである。
【００５６】
なお、色データの変換処理は、ＲＧＢ入力からＹＭＣＫ出力への多次元ＬＵＴを用いた色変換に限定されるものではない。たとえば、ある色軸（たとえばＬａｂ空間の各軸）のデータを同一軸上で異なるデータ値に変換するための１次元ＬＵＴを用いる処理であってもよい。ただし、１次元ＬＵＴであれば、１つの入力構成要素の色データから複数の出力構成要素の色データに変換するということはあり得ず、「出力色の複数を纏める」という概念が存在しないので、第１例の適用範囲は、入力構成要素数が“２”以上のものに限定される。
【００５７】
また、入力構成要素数が“２”である例としては、たとえば入力構成要素が３入力以上であるものについて、その内の２軸のみに着目した２次元ＬＵＴを用いた色変換が考えられる。具体的には、Ｌａｂ色空間で表される色データを入力とするケースにおいて、色相および彩度を表すａ，ｂのみに着目して、ａｂ空間の２次元ＬＵＴを用いて出力値ａ，ｂを求める色変換が考えられる。これは、明度については入力値を維持するが、色相および彩度についてデータ値を変える（色変換する）ことを意味する。この事例に前述の第１例の合成手法を適用する場合、出力値ａ，ｂについて合成すればよい。
【００５８】
また、入力構成要素数が“３”である例としては、ＲＧＢ，ｓＲＧＢ，Ｌａｂ，ＹＣｂＣｒ，ＹＭＣ，ＸＹＺなどの色空間データが考えられる。さらに、入力構成要素数が“４”である例としては、ＹＭＣＫ，Ｋ（墨量）＋Ｌａｂ，ＲＧＢ＋Ｌ（明度）などの色空間データが考えられる。また、ＲＧＢ入力からＹＭＣＫ出力への色変換のように、入力色空間と出力色空間とが同じものに限らず、異なるものであっても構わない。
【００５９】
以上のように、第１例のデータ合成の手法を適用して元のＬＵＴ格子点データの同一アドレスのデータを複数の出力色構成要素について纏めることでデータビット幅を大きくし、このデータビット幅の大きな状態でメモリにＬＵＴデータを格納することとすれば、ゲート規模（メモリサイズ）を小さくすることができ、メモリコストの低減を図ることができる。これにより、冗長性の低い、効率的なＬＵＴの構成方法とアクセス方法を提供することができる。
【００６０】
図３は、図２に示した第１例の合成手法を適用したデータ構造を持つＬＵＴを使用する場合における、第１実施形態のデータ変換装置１の基本的な構成例を示す図である。
【００６１】
第１実施形態のデータ変換装置１には、外部機器から、入力系の色成分Ｉ１，Ｉ２，…，Ｉｗの各Ｙ１ビットの入力デジタル色分解信号が入力される。データ変換装置１は、入力データを、出力系の色成分S1，S2，…，Sxの各Ｙ２ビットのデータに色変換する。たとえばＲ，Ｇ，Ｂの３色の各８ビットデータが入力されると、Ｙ，Ｍ，Ｃ，Ｋの４色の各８ビットのデータに色変換する。なお、この入力データや出力データの形式は、一例に過ぎず、他の色系列であってもよいし、ビット数Ｙ１，Ｙ２も“８”に限らない。
【００６２】
図示するように、第１実施形態のデータ変換装置１は、入力データに対応する色変換後の出力データを補間演算により求める際に使用される基礎データを格納するデータ格納部２と、データ格納部２との間データでやり取りをしながら、出力色ごとに、入力画素データに対応する出力画素データを補間演算により求める演算処理部４とを備える。
【００６３】
データ格納部２は、入力色信号要素で規定されるｗ次元格子空間（ｗ次元ＬＵＴ）の各格子点に対応づけられた格子点データに対応する基礎データが、図２にて説明したように、複数の出力色構成要素について所定の基準に従ってグループ化され、そのグループ内の同一アドレスの複数の基礎データが１つの合成データに纏められて格納されることで、ルックアップテーブルＬＵＴを構成するようになっている。たとえば、出力色の全てを纏めて１グループとする場合、補間演算に必要となる格子点データ数Ｔの分だけの参照用メモリ２T1，２T2，…，２Ttを使用する。各参照用メモリ２T1，２T2，…，２Ttには、出力色S1〜Sxについての格子点データが合成されて格納される。
【００６４】
これらメモリ２T1，２T2，…，２Ttは、１つの画素に対する色空間変換の処理時間として割り当てられている時間が非常に短時間である場合にも対応可能なように、それぞれ独立にアドレス設定可能かつ読出可能に、つまり独立に読出アクセス可能に構成されている。
【００６５】
演算処理部４は、信号処理系統として、出力系の色成分S1，S2，…，Sxのそれぞれに対応するように、S1信号生成部４S1，S2信号生成部４S2，…，Sx信号生成部４Sxを含む。また、演算処理部４は、データ格納部２に格納されている合成データを、出力構成要素ごとのデータに分離するデータ分離部１６を備える。ここで、第１実施形態のデータ分離部１６は、出力色S1，S2，…，Sxに対応して、個別回路が内部に設けられる。この演算処理部４は、たとえばＣＰＵ（Central Processing Unit ）やＤＳＰ（Digital Signal Processor）で構成するとよい。
【００６６】
データ格納部２と演算処理部４との間は、演算処理部４がデータ格納部２から基礎データを読み出すためにアドレス信号などの制御信号を送るアドレスバス６、およびデータ格納部２から読み出された基礎データを演算処理部４へ送るデータバス８で接続されている。
【００６７】
演算処理部４には、出力値を生成するための補間演算に用いられるデータが格納されているルックアップテーブルへのアクセスアドレスを生成するアドレス生成部が設けられる。そして、アドレスバス６では、データ格納部２内の個々のメモリ空間（メモリ２T1，２T2，…，２Tt）における基礎データの読出位置を設定するアドレス信号、クロック、あるいはチップセレクト信号などの伝送が行なわれる
データバス８は、データ変換前の入力データを図示しない外部装置から入力するためや、変換後の出力データを同一または別の外部装置へ出力するために用いられる。
【００６８】
このような構成において、演算処理部４は先ず、データ格納部２内のメモリ２T1，２T2，…，２Ttの任意のアドレスから、各出力色構成要素の格子点データであって複数の出力色構成要素について合成されたＮビットもしくはＮｉビットの合成データを読み出す。データ分離部１６は、読み出された合成データを、出力色構成要素ごとの格子点データに分離する。色成分S1，S2，…，Sxごとに設けられた信号生成部４S1〜４Sxは、分離された格子点データを用いて補完演算を行なうことで、色成分S1，S2，…，Sxごとに出力データを生成する。
【００６９】
図４は、第１実施形態のデータ変換装置１における色データの変換に関する処理手順の概要を示したフローチャートである。なお、以下の説明では、入力データが８ビットＲ，Ｇ，Ｂ、出力データが８ビットＹ，Ｍ，Ｃ，Ｋで、演算処理部４は、テトラ・ハイドラ補間演算をするものとして説明する。この場合、図３に示すように、演算処理部４には、Ｙ信号生成部４Ｙ，Ｍ信号生成部４Ｍ，Ｃ信号生成部４Ｃ，Ｋ信号生成部４Ｋを使用する。また、メモリ２には、４つのメモリ２T1，２T2，２T3，２T4を使用する。
【００７０】
第１実施形態の演算処理部４は、データ格納部２に格納されている色変換用のＬＵＴ（ルックアップテーブル）の格子点データを用いて、外部機器から入力された各信号（Ｒ，Ｇ，Ｂ）を、出力色（Ｙ，Ｍ，Ｃ，Ｋ）の色データに色変換する際に、テトラ・ハイドラ法を用いた補間処理を利用する。
【００７１】
このため、画像データＲ，Ｇ，Ｂが入力されると、演算処理部４は先ず、入力された８ビットのＲ，Ｇ，Ｂデジタル色分解信号を、画素ごとに、上位αビット（最上位ビット側から連続したα個）と下位βビット（最下位ビット側から連続したβ個）に分割する（Ｓ１００）。
【００７２】
次に、演算処理部４は、入力３要素の上位αビットのデータＲＨ，ＧＨ，ＢＨに基づいて、入力３要素のＲ，Ｇ，Ｂデータからなる３次元格子体（３次元ＬＵＴ）を構成したとき、各格子点で規定される多数の単位格子体（本例では単位立方体）のうちの何れの単位立方体に該当データが含まれるかを決定する（Ｓ１０２）。このステップＳ１００〜Ｓ１０２の処理を単位立方体判定処理という。
【００７３】
次に、演算処理部４は、該当する単位立方体の中をさらに６つの４面体に分割し（Ｓ１０４）、入力３要素の下位βビットのデータＲＬ，ＧＬ，ＢＬに基づいて何れの４面体に該当データが含まれるかを決定する（Ｓ１０６）。このステップＳ１０４〜Ｓ１０６の処理を４面体判定処理という。
【００７４】
次に、演算処理部４は、該当する４面体内での補間演算というテトラ・ハイドラ法により、入力値（入力された画素データ）に対応する出力値（出力すべき画素データ）を求める（Ｓ１０８）。
【００７５】
具体的には、図中右側に示すように、演算処理部４は、先ず、入力点を包含する基点の座標値に基づいて、同一アドレスのデータ群における補間演算に必要な４格子点を特定し抽出する４格子点特定処理（アクセスアドレス特定処理ともいう）を行なう（Ｓ１０８ａ）。つまり、演算処理部４は、４つのメモリ２T1，２T2，２T3，２T4について、テトラ・ハイドラ法による補間演算処理により出力値を生成するために用いられるデータが格納されているＬＵＴへのアクセスアドレスを特定する。
【００７６】
ＬＵＴの任意のアドレスには、出力構成要素数分のデータを合成してｎビット（本例では３２ビット）で格納されている。このため、演算処理部４のデータ分離部１６は、アクセスアドレス特定処理にて特定された入力値に対応する特定アドレスから格子点データを抽出し、この抽出した３２ビットの格子点データを、出力色構成要素の単位データであるｍビット（本例では８ビット）のデータに分離する（Ｓ１０８ｂ）。
【００７７】
演算処理部４は、４格子点特定処理（Ｓ１０８ａ）と並行して、該当する４面体について、４面体頂点から得られる４つの基礎データ（本例では元の格子点データそのもの）に対する重み係数（乗算係数）である補間係数を算出する補間係数算出処理を行なう（Ｓ１０８ｃ）。そして、演算処理部４は、該当する４面体について、４格子点特定処理（Ｓ１０８ｂ）により得られた基礎データ（格子点データ）と補間係数算出処理（Ｓ１０８ｃ）により得られた補間係数とに基づいて、入力値に対応する出力値Ｄｏｕｔを求める補間演算処理を行なう（Ｓ１０８ｄ）。
【００７８】
演算処理部４は、上記一連の処理を、出力色であるＹ，Ｍ，Ｃ，Ｋのそれぞれについて行なうとともに（Ｓ１１２）、全画素について行なうことで（Ｓ１１０）、入力されたＲＧＢ色空間の色データをＹＭＣＫ色空間のデータに変換する。
【００７９】
図５は、第１実施形態のデータ変換装置１において、ＲＧＢ入力からＹＭＣＫ出力への多次元ＬＵＴを用いた色変換を行なう場合のデータ構造の具体例を示す図である。図１８に示した従来例との比較のため、元の（従前の）ＬＵＴへの格納データ数を１７データ×１７データ×１７データ（４９１３個の格子点アドレスが必要）と規定する。
【００８０】
ここで、前述のように、演算処理部４は、テトラ・ハイドラ補間演算を行なうものとする。この場合、メモリ（データ格納部２）は、テトラ・ハイドラ補間演算に必要となる４つの頂点のデータをある単位時間内に同時に抽出できるように、４つの頂点用の４個のメモリ（ＲＡＭ０〜ＲＡＭ３）で構成される。
【００８１】
各頂点のアドレス位置には、ＹＭＣＫ４色分の格子点データが１つの合成データとして纏められて保持される。ここでは、ＲＧＢ入力データが８ビットであり、ＹＭＣＫ４色分が纏められることで、１ワードとしては、８ビット×４色＝３２ビットの状態で格納される。このような構成によれば、出力色ごとにメモリを個別に使用する場合よりも、特定用途向け半導体内に構成されるメモリとしてはその占有面積が格段に小さくなり、コスト面での経済的効果は絶大である。
【００８２】
図６は、第１実施形態のデータ変換装置１の具体的な構成を示すブロック図である。この図６に示すデータ変換装置１の構成は、Ｒ，Ｇ，Ｂの３次元入力の場合における図５に示したメモリ構成を適用した具体的な回路構成である。ここでも、入力デジタル色分解信号の構成要素がＲ，Ｇ，Ｂであり、それぞれ８ビットのデータであるとする。
【００８３】
図示するように、演算処理部４は、単位立方体判定処理（図４のステップＳ１００〜Ｓ１０２）およびアクセスアドレス特定処理（同Ｓ１０８ａ）を実行するアドレス生成部１０と、４面体判定処理（同Ｓ１０４〜Ｓ１０６）を実行する４面体判定部１４と、データ分離処理（同Ｓ１０８ｂ）を実行するデータ分離部１６と、補間係数算出処理（Ｓ１０８ｃ）を実行する補間係数生成部２０と、近傍４つのデータを用いるテトラ・ハイドラ補間演算処理（Ｓ１０８ｄ）を実行する出力色データ決定部２２とを備える。
【００８４】
データ格納部２は、テトラ・ハイドラ補間演算における４頂点に対応するように、第１頂点のデータを参照するためのＬＵＴを構成する第１頂点データ参照用メモリ２T1、第２頂点のデータを参照するためのＬＵＴを構成する第２頂点データ参照用メモリ２T2、第３頂点のデータを参照するためのＬＵＴを構成する第３頂点データ参照用メモリ２T3、および第４頂点のデータを参照するためのＬＵＴを構成する第４頂点データ参照用メモリ２T4を備えている。
【００８５】
それぞれのメモリ２T1〜２T4は同じ内容で同じ容量のＬＵＴを格納している。テトラ・ハイドラ補間演算を行なう場合、従来は図１８に示したように、同容量・同内容のメモリを４つ用意し、かつＹ，Ｍ，Ｃ，Ｋの４出力に応じて、計１６個のメモリを用意する必要があった。ここで、各頂点へのアドレスはそれぞれ異なるが、各色では同じ頂点のデータを必要としており、アクセスするメモリ実態は異なっていてもアドレス値は同じである。
【００８６】
よって、図３（Ａ）や図５に示したように、１つのアドレスの４つの出力値構成要素を纏めることで、メモリの簡素化を図ることが可能となる。この場合は出力値がそれぞれ８ビットのデータであり、出力色構成要素はＹ，Ｍ，Ｃ，Ｋの４色なので、１つのアドレスに３２ビットの合成データを格納することになる。
【００８７】
たとえば、第１頂点データ参照用メモリ２T1には４つの出力色Ｙ，Ｍ，Ｃ，Ｋの第１頂点についてのデータを纏めた合成データ“Ｙ１＋Ｍ１＋Ｃ１＋Ｋ１”（参照符号“１”が頂点番号を示す）が格納されている。同様に、第２頂点データ参照用メモリ２T2には合成データ“Ｙ２＋Ｍ２＋Ｃ２＋Ｋ２”が、第３頂点データ参照用メモリ２T3には合成データ“Ｙ３＋Ｍ３＋Ｃ３＋Ｋ３”が、第４頂点データ参照用メモリ２T4には合成データ“Ｙ４＋Ｍ４＋Ｃ４＋Ｋ４”が、それぞれ格納されている。
【００８８】
このようなメモリ構成にすると、ゲート規模の極小化の利点を享受することができる。前述のように、同容量のメモリを構成する場合、よりビット幅を大きくし、メモリの個数をより少なくした方が、ゲート規模を削減することができる。
【００８９】
第１実施形態のデータ分離部１６は、全出力色について纏められた合成データを各出力色成分に分離するものであり、４つの頂点用のメモリ２T1〜２T4に対応するように、第１出力色データ分離部１６T1、第２出力色データ分離部１６T2、第３出力色データ分離部１６T3、および第４出力色データ分離部１６T4を備えている。
【００９０】
出力色データ決定部２２は、４つの出力色Ｙ，Ｍ，Ｃ，Ｋに対応するように、Ｙ補間処理部２２Ｙ、Ｍ補間処理部２２Ｍ、Ｃ補間処理部２２Ｃ、およびＫ補間処理部２２Ｋを備えている。
【００９１】
アドレス生成部１０は、合成データ読出部の一例であって、単位格子体判定部の機能とアドレス設定部の両機能を備える。また、補間係数生成部２０と出力色データ決定部２２とにより、補間演算処理部が構成される。
【００９２】
図７は、演算処理部４における単位立方体判定処理（図２のステップＳ１００〜Ｓ１０２）を説明する図であって、出力値の代表値である格子点データを構成する元の３次元ＬＵＴと入力データとの関係を説明する図である。ここで図７（Ａ）は、ＬＵＴの３次元空間での概念図を示し、図７（Ｂ）は、各８ビットのＲＧＢ入力信号とそれぞれの信号の上位４ビット信号が対応する３次元ＬＵＴ上での位置関係を示す。
【００９３】
図７（Ａ）に示すように、Ｒ，Ｇ，Ｂの各軸における取り得る値の範囲を複数に分割する。ここでは各軸を１６に分割した例を示している。このような分割によって得られる格子点をアドレスとし、その格子点における変換後の値（代表値）を対応付けることによってＬＵＴを構成する。したがって、この例では、１７×１７×１７＝４９１３個の格子点について、それぞれ対応する変換後のデータを代表値として格納した３次元ＬＵＴを作成することができる。
【００９４】
演算処理部４は、入力された８ビットのＲ，Ｇ，Ｂデジタル色分解信号を上位αビット（最上位ビット側から連続したα個）と下位βビット（最下位ビット側から連続したβ個）に分割する。たとえば、演算処理部４は、α＝β＝４とし、入力されたＲ，Ｇ，Ｂを、それぞれ上位４ビットのＲＨ，ＧＨ，ＢＨと下位４ビットのＲＬ，ＧＬ，ＢＬに分割する。もちろん、α，βは、これ以外であってもかまわない。なお、“α＋β”は、入力データのビット数８と等しいことが好ましい。
【００９５】
たとえば、各ＬＵＴは、図７（Ａ）に示すように、ＲＧＢそれぞれの軸に対し、１７個のデータを保持している。ここで、図７（Ｂ）に示すように、上位４ビットデータの組合せ値で形成される単位立方体の中に、任意の入力Ｒ，Ｇ，Ｂの８ビットデータ（入力画素データ）の組合せ値の空間位置Ｐ２２が包含されるので、上位の４ビットデータＲＨ，ＧＨ，ＢＨで示す空間位置Ｐ２１を基点とした単位立方体（単位格子体）で入力８ビットデータを包含する単位立方体を特定することが可能となる。
【００９６】
単位立方体を構成する各頂点は、空間位置Ｐ２１を基準に生成可能である。すなわち、Ｐ２１（ＲＨ，ＧＨ，ＢＨ）を基点として、Ｐ２３（ＲＨ＋１，ＧＨ，ＢＨ），Ｐ２４（ＲＨ＋１，ＧＨ＋１，ＢＨ），Ｐ２５（ＲＨ，ＧＨ＋１，ＢＨ），Ｐ２６（ＲＨ，ＧＨ，ＢＨ＋１），Ｐ２７（ＲＨ＋１，ＧＨ，ＢＨ＋１），Ｐ２８（ＲＨ，ＧＨ＋１，ＢＨ＋１），Ｐ２９（ＲＨ＋１，ＧＨ＋１，ＢＨ＋１）が、単位立方体の頂点の座標となる。本実施形態の場合には、これらの頂点の座標は、元となる３次元ＬＵＴの格子点の座標に一致する。また、ＬＵＴへのアクセスアドレスの算出式は、以下の式（２）となる。
【数２】

【００９７】
演算処理部４は、この入力Ｒ，Ｇ，Ｂの８ビットデータの組合せ値に対応する出力値を、入力点近傍の４つの格子点データを使って、つまり、図７（Ｂ）に示す単位立方体の８頂点の座標のデータのうちの４つを基礎データとして使って、テトラ・ハイドラ法により求める。
【００９８】
図８は、演算処理部４における４面体判定処理（図４のステップＳ１０４〜Ｓ１０６）を説明する図であって、入力データが包含される単位立方体と、この単位立方体を６つに分割して得られる４面体との関係を示す図である。先にも述べたように、第１実施形態の演算処理部４は、入力デジタル色分解信号Ｒ，Ｇ，Ｂの組合せ値に対応する出力信号ＹＭＣＫのそれぞれを、データ格納部２に格納されている基礎データ（格子点データ）に基づいてテトラ・ハイドラ法による補間演算により求める。
【００９９】
図７（Ｂ）に示した単位立方体は、さらに図８に示すように、同体積の６つの単位４面体に分割可能である。入力Ｒ，Ｇ，Ｂの空間位置Ｐ２２がどの４面体に包含されるかは、下位のビットデータＲＬ，ＧＬ，ＢＬの大小関係から、図８下部に示す表２のように特定可能である。
【０１００】
たとえば、下位ビットデータＲＬ，ＧＬ，ＢＬがＲＬ≧ＧＬ≧ＢＬの関係であるとき、入力Ｒ，Ｇ，Ｂは（ＲＨ，ＧＨ，ＢＨ），（ＲＨ＋１，ＧＨ，ＢＨ），（ＲＨ＋１，ＧＨ＋１，ＢＨ），（ＲＨ＋１，ＧＨ＋１，ＢＨ＋１）を頂点とした４面体１に属する。ＲＬ＞ＢＬ＞ＧＬの関係であるとき、入力Ｒ，Ｇ，Ｂは（ＲＨ，ＧＨ，ＢＨ），（ＲＨ＋１，ＧＨ，ＢＨ），（ＲＨ＋１，ＧＨ，ＢＨ＋１），（ＲＨ＋１，ＧＨ＋１，ＢＨ＋１）を頂点とした４面体２に属する。ＢＬ≧ＲＬ＞ＧＬの関係であるときは、入力Ｒ，Ｇ，Ｂは（ＲＨ，ＧＨ，ＢＨ），（ＲＨ＋１，ＧＨ，ＢＨ＋１），（ＲＨ，ＧＨ，ＢＨ＋１），（ＲＨ＋１，ＧＨ＋１，ＢＨ＋１）を頂点とした４面体３に属する。
【０１０１】
同様に、ＢＬ＞ＧＬ≧ＲＬの関係であるときは、入力Ｒ，Ｇ，Ｂは（ＲＨ，ＧＨ，ＢＨ），（ＲＨ，ＧＨ，ＢＨ＋１），（ＲＨ，ＧＨ＋１，ＢＨ＋１），（ＲＨ＋１，ＧＨ＋１，ＢＨ＋１）を頂点とした４面体４に属する。ＧＬ≧ＢＬ＞ＲＬの関係であるときは、入力Ｒ，Ｇ，Ｂは（ＲＨ，ＧＨ，ＢＨ），（ＲＨ，ＧＨ＋１，ＢＨ），（ＲＨ，ＧＨ＋１，ＢＨ＋１），（ＲＨ＋１，ＧＨ＋１，ＢＨ＋１）を頂点とした４面体５に属する。ＧＬ＞ＲＬ≧ＢＬの関係であるときは、入力Ｒ，Ｇ，Ｂは（ＲＨ，ＧＨ，ＢＨ），（ＲＨ，ＧＨ＋１，ＢＨ），（ＲＨ＋１，ＧＨ＋１，ＢＨ），（ＲＨ＋１，ＧＨ＋１，ＢＨ＋１）を頂点とした４面体６に属する。
【０１０２】
そこで、先ず演算処理部４のアドレス生成部１０は、画像データＲ，Ｇ，Ｂが入力されると、その入力データＲ，Ｇ，Ｂを上位ビットデータＲＨ，ＧＨ，ＢＨ（本例では上位４ビット；纏めてＤＨとする）と下位ビットデータＲＬ，ＧＬ，ＢＬ（本例では下位４ビット；纏めてＤＬとする）に分割し、その内の上位４ビットのデータＤＨに基づいて、図７（Ｂ）に示したように、３次元ＬＵＴにおける入力点が包含される単位立方体の入力位置Ｐ２２を特定する。また、アドレス生成部１０は、分割したデータの内の下位ビットデータＤＬを４面体判定部１４と補間係数生成部２０とに入力する。
【０１０３】
４面体判定部１４は、アドレス生成部１０で生成された入力ＲＧＢ信号の各下位４ビット信号ＤＬを受け取ると、その３つの信号の大小関係を比較演算することで、入力の各８ビット信号が単位立方体内のどの４面体に包含されるかを判定する。４面体判定部１４は、判定結果をアドレス生成部１０と補間係数生成部２０に通知する。
【０１０４】
４面体判定部１４における比較条件は、図８の４面体構成に則した場合、入力ＲＧＢ信号の下位４ビット信号ＤＬをそれぞれＲＬ，ＧＬ，ＢＬとすると、前述のように、ＲＬ≧ＧＬ≧ＢＬのときは４面体１、ＲＬ＞ＢＬ＞ＧＬのときは４面体２、ＢＬ≧ＲＬ＞ＧＬのときは４面体３、ＢＬ＞ＧＬ≧ＲＬのときは４面体４、ＧＬ≧ＢＬ＞ＲＬのときは４面体５、ＧＬ＞ＲＬ≧ＢＬのときは４面体６となる。
【０１０５】
図９は、基点Ｐ２１と４面体判定部１４による判定結果に基づく４頂点の座標位置との関係を示す図である。図中ａが図７（Ｂ）に示した基点Ｐ２１に対応する。
【０１０６】
入力データ位置を示す入力点Ｐ２２は、入力データの上位ビットデータＲＨ，ＧＨ，ＢＨで規定される単位立方体に包含され、この単位立方体の位置が、図７（Ｂ）の基点Ｐ２１（つまり図９のａ）の座標値により特定される。
【０１０７】
アドレス生成部１０は、４面体判定部１４による４面体判定処理結果を受けて、入力点Ｐ２２を包含する単位立方体の８頂点（図９のａ〜ｈ）のうち、４面体判定部１４による判定結果に基づく４頂点アドレスを式（２）に従って算出し、４つのメモリ２T1，２T2，２T3，２T4に設定する。こうすることで、データ格納部２の各メモリ２T1，２T2，２T3，２T4からのデータ読出し時には、設定されたアドレスから、入力点Ｐ２２近傍の４頂点の格子点データを示す３２ビットの合成データがデータ分離部１６に入力されるようになる。
【０１０８】
メモリ２T1〜２T4から、第１頂点〜第４頂点のデータが、ある単位時間内に同時に出力される。この出力されたデータは、Ｙ，Ｍ，Ｃ，Ｋの４つの同一頂点（同一アドレス）のデータが一体になった合成データである。このため、データ分離部１６の各出力色データ分離部１６T1〜１６T4は、予め定められているデータ合成手法を参照して、出力色Ｙ，Ｍ，Ｃ，Ｋの単位データ、すなわち８ビット単位のデータに分離する。この８ビット単位に分離されたデータを補間対象データＤ１，Ｄ２，Ｄ３，Ｄ４という。個々の出力色ごとに纏まった形式で合成データが用意されているので、データ分離処理が簡易である。
【０１０９】
図９中の表３から分かるように、６つの４面体ごとに補間対象データＤ１，Ｄ２，Ｄ３，Ｄ４に割り当てられる格子点が異なる。データ分離部１６は、分離した各出力色についての補間対象データＤ１，Ｄ２，Ｄ３，Ｄ４を、対応する補間処理部２２Ｙ〜２２Ｋに入力する。
【０１１０】
図１０は、演算処理部４における出力値算出処理（図４のステップＳ１０８）の補間係数算出処理（Ｓ１０２ｃ）と補間演算処理（Ｓ１０２ｄ）を説明する図である。ここで、図１０（Ａ）は補間演算にて必要となる補間係数の算出式を示す図表であり、図１０（Ｂ）は補間演算処理の機能ブロック図である。
【０１１１】
補間係数算出処理（Ｓ１０２ｃ）は、補間係数生成部２０が実行する。補間係数生成部２０は、アドレス生成部１０から入力された入力ＲＧＢ信号の下位４ビット信号ＤＬ（ＲＬ，ＧＬ，ＢＬ）と４面体判定部１４にて定まった４面体判定結果とに基づいて補間演算係数Ｗ１，Ｗ２，Ｗ３，Ｗ４を算出する。図１０（Ａ）示す表４は、テトラ・ハイドラ法による補間演算で必要となる補間係数の算出式を示している。表４から分かるように、６つの４面体ごとに計算式が異なる。
【０１１２】
補間係数生成部２０は、４面体判定処理（図４のステップＳ１０４〜Ｓ１０６）の結果を受けて、入力値が包含される４面体ごとに定められた計算式を表３に基づき切り替える。そして、４面体を構成する４格子点の下位ビットデータＲＬ，ＧＬ，ＢＬに基づいて補間係数Ｗ１，Ｗ２，Ｗ３，Ｗ４を算出し、補間係数Ｗ１，Ｗ２，Ｗ３，Ｗ４を補間処理部２２Ｙ〜２２Ｋに共通に入力する。この補間係数算出処理は、従来のデータ変換装置における処理と同様である。
【０１１３】
補間演算処理（Ｓ１０２ｄ）は、出力色データ決定部２２が実行する。補間処理部２２Ｙ〜２２Ｋのそれぞれは、補間係数生成部２０から入力された補間係数Ｗ１，Ｗ２，Ｗ３，Ｗ４と対応するデータ分離部１６T1〜１６T4から入力された補間対象データＤ１，Ｄ２，Ｄ３，Ｄ４（４頂点の基礎データ）とに基づいて、最終出力値Ｄｏｕｔを算出するというテトラ・ハイドラ法による補間演算処理を行なう。たとえば、式（３）に従って、最終出力値Ｄｏｕｔを算出する。式（３）を機能的に示したものが図１０（Ｂ）である。なお、式（３）では１次式補間で出力値を求めているが、これに限らず、２次式やその他の高次の補間演算式を利用してもよい。
【数３】

【０１１４】
図１１は、メモリデータ構造の第２例を説明する図である。第１例のデータ構造は、出力色構成要素のうちの一部もしくは全てのデータを纏めて（合成して）格納するものであった。これに対して、第２例のデータ構造は、複数の画像属性のうちの一部もしくは全てのデータを纏めて（合成して）格納する点に特徴を有する。以下、第１例のデータ構造と異なる点についてのみ説明する。
【０１１５】
第１例のデータ構造では、入力値に対応するＬＵＴ内の近傍データは各頂点において１つであり、補間演算されて出力されるデータは１つの入力に対して１対１に定まるものであった。しかし、プリンタなどのハードコピー出力装置はその出力文書中の構成要素（たとえば、テキスト、コンピュータ・グラフィックス画像、写真画像など）ごとに異なった色変換を行なうことが望まれる。つまり、各頂点のデータとして、それぞれ異なる画像属性に応じた複数のデータが必要となるということである。
【０１１６】
この場合、従来の構成では、１つの画像属性についての必要メモリ数（補間演算に必要となるＬＵＴの格子点数（頂点数）×出力色構成要素数Ｘ）を、必要とする画像属性の数の分だけ用意しなければならない。たとえば、ＲＧＢ入力色データをテトラ・ハイドラ補間演算によりＹＭＣＫ出力色データに変換するためには、Ｙへの変換に４つの格子点抽出用として４個、同様にＭへの変換用に４個、Ｃへの変換用に４個、Ｋへの変換用に４個の計１６個が必要であるから、画像属性の数をＧとすれば、１６・Ｇ個のメモリが必要となりメモリ回路の規模が増大する。
【０１１７】
そこで、第２例のデータ構造においては、先ず従来例と同様に、１つの画像属性についての必要メモリ数である、補間演算に必要となるＬＵＴの格子点数（頂点数）×出力色構成要素数Ｘと等しい数のメモリで構成する。ＲＧＢ入力色データをテトラ・ハイドラ補間演算によりＹＭＣＫ出力色データに変換する場合、４個×４色＝１６個メモリを使用する。
【０１１８】
一方、新たに構成するＬＵＴの各格子点のアドレス位置には、４つの画像属性に対応した例を示す図１１のように、出力色構成要素のそれぞれについて、テキスト、コンピュータ・グラフィックス画像、あるいは写真画像などの画像属性に応じた選択データ構成要素のうちの一部もしくは全てのデータを纏めて（合成して）格納する。図１１（Ａ）に示すように全てを纏めた方が、図１１（Ｂ）に示すように一部だけを纏めるよりも、ビット幅が広がるのはいうまでもない。
【０１１９】
ここで、合成される任意の格子点のデータビット幅をＬビット、１つの構成要素のデータビット幅をＭビット（Ｌ，Ｍは正の整数）、元の（従前の）ＬＵＴの任意アドレスに格納されているデータサイズをＹビット（Ｙは正の整数）、画像属性に応じた選択データ構成要素数をＺ（Ｚは２以上の正の整数）、選択データ構成要素数（Ｚ）を所定数ｎのグループに分割（Ｚ＝ｚ１＋ｚ２＋…＋ｚｎ；ｚｊは正の整数）することとすれば、各グループのＬＵＴ格子点のデータビット幅ＬやＬｊ（単位はビット）は、以下の式（４）で示すことができる。
【数４】

【０１２０】
Ｚｊ＝Ｚであれば、全てを纏めることを意味し、１グループのみとなる。なお、従前の例から分かるように、通常は、１つの構成要素のデータビット幅ＭとＬＵＴのデータサイズＹとは等しい。
【０１２１】
また式（４）を満足する限り、図１１（Ｂ）や図１１（Ｃ）に示すように、複数のグループに分けて纏めることもできる。また、図１１（Ｃ）に示すように、纏め方によっては、１構成要素のデータビット幅そのもので格納されるグループができる。
【０１２２】
図１１（Ａ）〜図１１（Ｃ）に示すように、複数の画像属性を纏めるデータ合成の手法には種々の手法があるが、これらは、出力色構成要素のそれぞれについて、同じデータ合成の手法を採用することが好ましい。ただし、出力色構成要素に応じて、データ合成の手法を変えることを排除するものではない。たとえば、出力色Ｙ，Ｍ，Ｃについては図１１（Ａ）の手法を採用しつつ、出力色Ｋについては図１１（Ｂ）の手法を採用するなどである。
【０１２３】
なお、複数のグループに分けるということから、この場合、選択データ構成要素数Ｚが３以上であることが前提となる。纏めるからには、選択データ構成要素数Ｚが複数（２以上）でなければならず、Ｚ＝２の場合、これを纏めると１つのグループしかできなからである。
【０１２４】
この第２例は、第１例とは異なり、ある色軸（たとえばＬａｂ空間の各軸）のデータを同一軸上で異なるデータ値に変換するための１次元ＬＵＴを用いる処理、すなわち入力構成要素数が１のものにも適用可能である。画像属性に応じて１次元ＬＵＴの変換特性を切り替えるということが存在するからである。
【０１２５】
入力構成要素数が“２”、“３”、あるいは“４”である例については、第１例の場合と同様である。色空間の具体例についても第１例の場合と同様であるので、その説明は割愛する。
【０１２６】
図１２は、図１１に示した第２例の合成手法を適用したデータ構造を持つＬＵＴを使用する場合における、第２実施形態のデータ変換装置１の基本的な構成例を示す図である。図３に示した第１実施形態の構成との違いは、データ格納部２におけるメモリデータ構造と、データ分離部１６の機能である。以下、相違点について説明する。
【０１２７】
第２実施形態のデータ格納部２は、図１１にて説明したように、複数の画像属性に対応した選択データ構成要素について所定の基準に従ってグループ化され、そのグループ内の同一アドレスの複数の基礎データが１つの合成データに纏められて格納されることで、ルックアップテーブルＬＵＴを構成する。たとえば、全ての選択データ構成要素を纏めて１グループとする場合、補間演算に必要となるＬＵＴの格子点数（頂点数）Ｔ×出力色構成要素数Ｘと等しい数Ｔ×Ｘの分だけの参照用メモリ２S1T1，２S1T2，…，２S1Tt，２S2T1，２S2T2，…，２S2Tt，２SxT1，２S1T2，…，２SxTtを使用する。たとえば、ＲＧＢ入力色データをテトラ・ハイドラ補間演算によりＹＭＣＫ出力色データに変換する場合、Ｔ＝４，Ｘ＝４であるから１６個のメモリを使用する。各参照用メモリ２T1，２T2，…，２Ttには、画像属性Z1〜Zzについての格子点データが合成されて格納される。
【０１２８】
これらメモリ２S1T1，２S1T2，…，２SxTtは、１つの画素に対する色空間変換の処理時間として割り当てられている時間が非常に短時間である場合にも対応可能なように、それぞれ独立にアドレス設定可能かつ読出可能に、つまり独立に読出アクセス可能に構成されている。
【０１２９】
演算処理部４は、信号処理系統として、出力系の色成分S1，S2，…，Sxのそれぞれに対応するように、S1信号生成部４S1，S2信号生成部４S2，…，Sx信号生成部４Sxを含む。また、演算処理部４は、データ格納部２に格納されている合成データを、テキスト、コンピュータ・グラフィックス画像、あるいは写真画像などの画像属性（選択データ構成要素）ごとのデータに分離するデータ分離部１６を備えている。ここで、第２実施形態のデータ分離部１６は、補間演算に必要となるＬＵＴの格子点T1，T2，…，Tt並びに出力色S1，S2，…，Sxに対応して、個別回路が内部に設けられる。
【０１３０】
演算処理部４は先ず、データ格納部２内のメモリ２S1T1，２S1T2，…，２S1Tt，２S2T1，２S2T2，…，２S2Tt，２SxT1，２S1T2，…，２SxTtの任意のアドレスから、各出力色構成要素の格子点データであって複数の画像属性について合成されたＬビットもしくはＬｊビットの合成データを読み出す。データ分離部１６は、予め定められているデータ合成手法を参照して、読み出された合成データの中から、テキスト、コンピュータ・グラフィックス画像、あるいは写真画像などの画像属性ごとの格子点データに分離し、処理対象の画像属性に応じた格子点データを選択する。個々の画像属性ごとに纏まった形式で合成データが用意されているので、データ分離・選択処理が簡易である。色成分S1，S2，…，Sxごとに設けられた信号生成部４S1〜４Sxは、分離・選択された格子点データを用いて補完演算を行なうことで、色成分S1，S2，…，Sxごとに出力データを生成する。
【０１３１】
なお、第２実施形態のデータ変換装置１における色データの変換に関する処理手順は、データ分離部１６におけるデータ分離処理が画像属性に応じた分離・選択の処理となる点を除いて、第１実施形態の処理手順と同様である。処理手順を示したフローチャートは、図示を割愛する。
【０１３２】
第２実施形態の構成によれば、前述のように、データ格納部２内のメモリ２S1T1，２S1T2，…，２S1Tt，２S2T1，２S2T2，…，２S2Tt，２SxT1，２S1T2，…，２SxTtの各頂点のアドレス位置には、複数の画像属性の格子点データが１つの合成データとして纏められて保持される。つまり、出力文書中の構成要素（画像オブジェクト）ごとに異なった色変換を行なうために、１つの画像属性に応じただけのメモリの使用個数（補間演算に必要となるＬＵＴの格子点数（頂点数）Ｔ×出力色構成要素数Ｘ）に維持しつつ、データビット幅を広くすることで全容量を維持する。このような構成によれば、出力文書中の画像オブジェクトごとに異なった色変換を行なうために全メモリ容量が増加する場合であっても、画像オブジェクトごとにメモリを個別に使用する場合よりも、特定用途向け半導体内に構成されるメモリとしてはその占有面積が格段に小さくなり、コスト面での経済的効果は絶大である。
【０１３３】
＜＜メモリデータの第３の構成例と対応回路＞＞
図１３は、メモリデータ構造の第３例を説明する図である。この第３例は、第１例と第２例とを組み合わせたものである。
【０１３４】
この第３例のデータ構造は、先ず第１例と同様に、補間演算に必要となるデータをある単位時間内に同時に抽出できるように、補間演算に必要となるＬＵＴの格子点数（頂点数）と等しい数（Ｔ個；テトラ・ハイドラ補間の場合は４個）のメモリで構成する。
【０１３５】
一方、新たに構成するＬＵＴの各格子点のアドレス位置には、４つの出力色および３つの画像属性に対応した例を示す図１３のように、出力色構成要素（本例ではＹ，Ｍ，Ｃ，Ｋ）のうちの一部もしくは全てのデータを纏めるとともに、複数の画像属性のうちの一部もしくは全てのデータを纏めて（合成して）格納する。図１３（Ａ）に示すように、複数の出力色構成要素並びに複数の画像属性の全てについて纏めた方が、図１３（Ｂ）に示すように一部だけを纏めるよりも、ビット幅が広がるのはいうまでもない。
【０１３６】
ここで、合成される任意の格子点のデータビット幅をＫビット、１つの構成要素のデータビット幅をＭビット（Ｌ，Ｍは正の整数）、元の（従前の）ＬＵＴの任意アドレスに格納されているデータサイズをＹビット（Ｙは正の整数）、出力色の構成要素数をＸ（Ｘは２以上の正の整数）、出力色構成要素（Ｘ）を所定数ｎのグループに分割（Ｘ＝ｘ１＋ｘ２＋…＋ｘｎ；ｘｉは正の整数）し、また画像属性に応じた選択データ構成要素数をＺ（Ｚは２以上の正の整数）、選択データ構成要素数（Ｚ）を所定数ｎのグループに分割（Ｚ＝ｚ１＋ｚ２＋…＋ｚｎ；ｚｊは正の整数）することとすれば、各グループのＬＵＴ格子点のデータビット幅Ｋ，Ｋｉ，Ｋｊ，Ｋｉｊ（単位はビット）は、以下の式（５）で示すことができる。
【数５】

【０１３７】
ｘｉ＝ＸかつＺｊ＝Ｚであれば、全てを纏めることを意味し、１グループのみとなる。Ｘｉ＝ＸかつＺｊ≠Ｚのときは、出力色構成要素については１グループのみとなる。Ｘｉ≠ＺかつＺｊ＝Ｚのときは、画像属性については１グループのみとなる。
【０１３８】
また第１例や第２例で述べたと同様に、式（５）を満足する限り、出力構成要素および画像属性（選択データ構成要素）の少なくとも一方について複数のグループに分けて纏めることもできる。なお、複数のグループに分けるということから、この場合、出力構成要素数Ｘおよび選択データ構成要素数Ｚの少なくとも一方が３以上であることが前提となる。纏めるからには、出力構成要素数Ｘあるいは選択データ構成要素数Ｚが複数（２以上）でなければならず、Ｘ＝２あるいはＺ＝２の場合、これを纏めると１つのグループしかできなからである。
【０１３９】
なお、この第３例は、第２例とは異なり、入力構成要素数が“２”以上のものに限定される。第１例と組み合わせたものであるからである。入力構成要素数が“３”、あるいは“４”である例については、第１例の場合と同様である。色空間の具体例についても第１例の場合と同様であるので、その説明は割愛する。
【０１４０】
図１４は、図１３に示した第３例の合成手法を適用したデータ構造を持つＬＵＴを使用する場合における、第３実施形態のデータ変換装置１の基本的な構成例を示す図である。図３に示した第１実施形態あるいは図１２に示した第２実施形態の構成との違いは、データ格納部２におけるメモリデータ構造と、データ分離部１６の機能である。以下、相違点について説明する。
【０１４１】
第３実施形態のデータ格納部２は、図１３にて説明したように、複数の出力色構成要素並びに複数の画像属性に対応した選択データ構成要素について所定の基準に従ってグループ化され、そのグループ内の同一アドレスの複数の基礎データが１つの合成データに纏められて格納されることで、ルックアップテーブルＬＵＴを構成する。たとえば、全ての出力色構成要素並びに全ての選択データ構成要素を纏めて１グループとする場合、第１例と同様に、補間演算に必要となる格子点データ数Ｔの分だけの参照用メモリ２T1，２T2，…，２Ttを使用する。各参照用メモリ２T1，２T2，…，２Ttには、画像属性Z1〜Zz並びに出力色S1〜Sxについての格子点データが合成されて格納される。
【０１４２】
これらメモリ２T1，２T2，…，２Ttは、１つの画素に対する色空間変換の処理時間として割り当てられている時間が非常に短時間である場合にも対応可能なように、それぞれ独立にアドレス設定可能かつ読出可能に、つまり独立に読出アクセス可能に構成されている。
【０１４３】
演算処理部４は、信号処理系統として、出力系の色成分S1，S2，…，Sxのそれぞれに対応するように、S1信号生成部４S1，S2信号生成部４S2，…，Sx信号生成部４Sxを含む。また、演算処理部４は、データ格納部２に格納されている合成データを、テキスト、コンピュータ・グラフィックス画像、あるいは写真画像などの画像属性（選択データ構成要素）ごとのデータに分離するとともに出力構成要素ごとのデータに分離するデータ分離部１６を備えている。ここで、第３実施形態のデータ分離部１６は、出力色S1，S2，…，Sxに対応して、個別回路が内部に設けられる。
【０１４４】
演算処理部４は先ず、データ格納部２内のメモリ２T1，２T2，…，２Ttの任意のアドレスから、各出力色構成要素の格子点データであって複数の出力色構成要素並びに複数の画像属性について合成されたＫビットもしくはＫｉビットの合成データを読み出す。データ分離部１６は、予め定められているデータ合成手法を参照して、読み出された合成データの中から、テキスト、コンピュータ・グラフィックス画像、あるいは写真画像などの画像属性ごとの格子点データに分離し処理対象の画像属性に応じた格子点データを選択するとともに出力色構成要素ごとの格子点データに分離する。個々の出力色および画像属性ごとに纏まった形式で合成データが用意されているので、データ分離・選択処理が簡易である。なお、処理対象の画像属性に応じた格子点データの選択処理（画像属性選択処理）と出力色構成要素ごとの格子点データへの分離処理（出力色分離処理）とは、その順序を問わない。
【０１４５】
色成分S1，S2，…，Sxごとに設けられた信号生成部４S1〜４Sxは、出力色構成要素ごとに分離された格子点データを用いて補完演算を行なうことで、色成分S1，S2，…，Sxごとに出力データを生成する。
【０１４６】
なお、第３実施形態のデータ変換装置１における色データの変換に関する処理手順は、データ分離部１６におけるデータ分離処理として画像属性に応じた分離・選択の処理が追加される点を除いて、第１実施形態の処理手順と同様である。処理手順を示したフローチャートは、図示を割愛する。
【０１４７】
図１５は、第３実施形態のデータ変換装置１において、ＲＧＢ入力からＹＭＣＫ出力への多次元ＬＵＴを用いた色変換を行なう場合のデータ構造の具体例を示す図である。２つの画像属性並びに４つの出力色ＹＭＣＫの事例で示す。図１８に示した従来例との比較のため、元の（従前の）ＬＵＴへの格納データ数を１７データ×１７データ×１７データ（４９１３個の格子点アドレスが必要）と規定する。
【０１４８】
ここで、前述のように、演算処理部４は、テトラ・ハイドラ補間演算を行なうものとする。この場合、メモリ（データ格納部２）は、テトラ・ハイドラ補間演算に必要となる４つの頂点のデータをある単位時間内に同時に抽出できるように、４つの頂点用の４個のメモリ（ＲＡＭ０〜ＲＡＭ３）で構成される。
【０１４９】
各頂点のアドレス位置には、２つの画像属性並びにＹＭＣＫ４色分の格子点データが１つの合成データとして纏められて保持される。ここでは、ＲＧＢ入力データが８ビットであり、２つの画像属性並びＹＭＣＫ４色分が纏められることで、１ワードとしては、８ビット×２×４色＝６４ビットの状態で格納される。つまり、出力文書中の構成要素（画像オブジェクト）ごとに異なった色変換を行なうために、補間演算に必要となるＬＵＴの格子点数（頂点数）Ｔに維持しつつ、データビット幅を広くすることで全容量を維持する。このような構成によれば、画像属性並びに出力色ごとにメモリを個別に使用する場合よりも、特定用途向け半導体内に構成されるメモリとしてはその占有面積が格段に小さくなり、コスト面での経済的効果は絶大である。
【０１５０】
図１６は、第３実施形態のデータ変換装置１の具体的な構成を示すブロック図である。この図１６に示すデータ変換装置１の構成は、Ｒ，Ｇ，Ｂの３次元入力の場合における図５に示したメモリ構成を適用した具体的な回路構成である。ここでも、入力デジタル色分解信号の構成要素がＲ，Ｇ，Ｂであり、それぞれ８ビットのデータであるとする。
【０１５１】
第３実施形態のデータ変換装置１は、アドレス生成部１０、４面体判定部１４、補間係数生成部２０、および出力色データ決定部２２の機能が、図６に示した第１実施形態のデータ変換装置１と同様である。一方、データ分離部１６の構成が第１実施形態のデータ変換装置１と異なる。
【０１５２】
具体的には、第３実施形態のデータ分離部１６は、全画像属性並びに全出力色について纏められた合成データを各出力色成分に分離するものであり、先ず４つの頂点用のメモリ２T1〜２T4に対応するように、第１データ選択部１７T1、第２データ選択部１７T2、第３データ選択部１７T3、および第４データ選択部１７T4と、第１出力色データ分離部１６T1、第２出力色データ分離部１６T2、第３出力色データ分離部１６T3、および第４出力色データ分離部１６T4とを備える。
【０１５３】
データ分離部１６の各データ選択部１７T1〜１７T4は、メモリ２T1〜２T4から出力された合成データの中から、テキスト、コンピュータ・グラフィックス画像、あるいは写真画像などの画像属性ごとの格子点データに分離し処理対象の画像属性に応じた格子点データを選択して、対応する出力色データ分離部１６T1〜１６T4に入力する。各出力色データ分離部１６T1〜１６T4は、出力色Ｙ，Ｍ，Ｃ，Ｋの単位データ、すなわち８ビット単位の補間対象データＤ１，Ｄ２，Ｄ３，Ｄ４に分離する。データ分離部１６は、分離した各出力色についての補間対象データＤ１，Ｄ２，Ｄ３，Ｄ４を、対応する補間処理部２２Ｙ〜２２Ｋに入力する。
【０１５４】
なお、図１４にて説明したように、画像属性選択処理と出力色分離処理とは順序を問わないので、各データ選択部１７T1〜１７T4と各出力色データ分離部１６T1〜１６T4との配列順を逆にしてもよい。
【０１５５】
第１実施形態のデータ変換装置１においては、入力値に対応するＬＵＴ内の近傍データは各頂点において１つであり、補間演算されて出力されるデータは１つの入力に対して１対１に定まるものであった。しかし、プリンタなどのハードコピー出力装置はその出力文書中の構成要素（たとえば、テキスト、コンピュータ・グラフィックス画像、写真画像など）ごとに異なった色変換を行なうことが望まれる。つまり、各頂点のデータとして複数のデータが必要となるということである。ある頂点のデータのうちどのデータを適用して後段の補間演算を行なうかは、たとえば入力の画像データに付加データ（タグ；ｔａｇ）を付加することで実現する。
【０１５６】
アドレス生成部１０によりアドレスされたデータ格納部２内の各メモリ２T1〜２T4は、Ｙ成分の２つのデータ（Ｙ０，Ｙ１）、Ｍ成分の２つのデータ（Ｍ０，Ｍ１）、Ｃ成分の２つのデータ（Ｃ０，Ｃ１）、Ｋ成分の２つのデータ（Ｋ０，Ｋ１）の合計８つのデータを一塊りの合成データとして出力する。ここで、ビット幅は、元のＬＵＴ格子点データが８ビットなので、６４ビットである。
【０１５７】
入力ＲＧＢのデータは、１画素ごとに、前述した文書中の構成要素を示すタグデータを持つ。ここでは１ビットのデータであるとする。タグデータは各データ選択部１７T1〜１７T4に入力される。タグデータが“０”である場合、各データ選択部１７T1〜１７T4は、６４ビットの合成データの中から、第１の画像属性の格子点データＹ０，Ｍ０，Ｃ０，Ｋ０を纏めた上位３２ビット部分のデータを分離・選択して、対応する出力色データ分離部１６T1〜１６T4に渡す。
【０１５８】
一方、タグデータが“１”である場合、各データ選択部１７T1〜１７T4は、６４ビットの合成データの中から、第２の画像属性の格子点データＹ１，Ｍ１，Ｃ１，Ｋ１を纏めた下位位３２ビット部分のデータを分離・選択して、対応する出力色データ分離部１６T1〜１６T4に渡す。各出力色データ分離部１６T1〜１６T4は、出力色Ｙ，Ｍ，Ｃ，Ｋの補間対象データＤ１，Ｄ２，Ｄ３，Ｄ４に分離し、分離した各出力色についての補間対象データＤ１，Ｄ２，Ｄ３，Ｄ４を、対応する補間処理部２２Ｙ〜２２Ｋに入力する。
【０１５９】
補間処理部２２Ｙ〜２２Ｋには、補間係数生成部２０から補間係数Ｗ１，Ｗ２，Ｗ３，Ｗ４が共通に入力される。補間係数Ｗ１，Ｗ２，Ｗ３，Ｗ４は入力点が包含される４面体ごとに定まるのでタグデータの状態には影響を受けない。補間処理部２２Ｙ〜２２Ｋのそれぞれは、第１実施形態と同様に、補間係数Ｗ１，Ｗ２，Ｗ３，Ｗ４と対応する補間対象データＤ１，Ｄ２，Ｄ３，Ｄ４とを使用して、テトラ・ハイドラ法による補間演算処理を行なうことで、対応する出力色のデータを生成する。
【０１６０】
以上、本発明を実施形態を用いて説明したが、本発明の技術的範囲は上記実施形態に記載の範囲には限定されない。発明の要旨を逸脱しない範囲で上記実施形態に多様な変更または改良を加えることができ、そのような変更または改良を加えた形態も本発明の技術的範囲に含まれる。
【０１６１】
また、上記の実施形態は、クレーム（請求項）にかかる発明を限定するものではなく、また実施形態の中で説明されている特徴の組合せの全てが発明の解決手段に必須であるとは限らない。前述した実施形態には種々の段階の発明が含まれており、開示される複数の構成要件における適宜の組合せにより種々の発明を抽出できる。実施形態に示される全構成要件から幾つかの構成要件が削除されても、効果が得られる限りにおいて、この幾つかの構成要件が削除された構成が発明として抽出され得る。
【０１６２】
たとえば、上記各実施形態では、データ分離処理を簡易にするために、個々の出力色あるいは画像属性ごとに纏まった形式で合成データを用意しておいたが、元の格子点データを各出力色や各画像属性に関わらずバラバラにして合成してもかまわない。データ変換処理時には、その合成手法を参照して元の格子点データを抽出すればよい。
【０１６３】
また、上述した実施形態では、画像属性や出力色の観点から複数の格子点データを合成した合成データを使用することとしていたが、これらに限らず、他の観点から取り決めた基準に従って複数の格子点データを合成した合成データを使用しても構わない。
【０１６４】
なお、本願出願人は、特願２００２−２７３９７６号にて、補間演算に必要となる格子点数（補間演算所用データ数）よりも少ない数のメモリ資源で色データ変換を実現する技術を提案している。元のルックアップテーブルの各格子点データを所定の基準に従って合成することで得られる合成データを新たなルックアップテーブルの格子点データとして格納する上述した本実施形態の手法を、この特願２００２−２７３９７６号に記載の技術における格子点データに適用することもできる。
【０１６５】
特願２００２−２７３９７６号に記載の技術は、多面体内での補間演算を行なう際に、補間演算所用データ数よりも少ない数のそれぞれ独立に制御可能なメモリ空間を使用し、そのメモリ空間のそれぞれについて、ｋ色の入力色信号要素で規定されるｋ次元格子空間の各格子点に対応づけられた格子点データに対応する基礎データを、所定の手順に従って１つのアドレスに複数の基礎アドレスが割り当てられるように対応付けて（パッキングして）格納し、同一アドレスに割り当てられた複数の基礎データの中から、処理対象の画素データに対応する補間演算所用データ数分の基礎データを抽出して、多面体内での補間演算を行なうというものである。
【０１６６】
特願２００２−２７３９７６号に記載の技術に、上述した実施形態を適用すれば、メモリサイズを低減することができることに加えて、補間演算所用データ数（４点補間の場合は４つ）未満という少ないメモリ資源で色データ変換を実現することができ、また消費電力や放射ノイズを低減することもできる。
【０１６７】
【発明の効果】
以上の説明から明らかなように、本発明においては、元のルックアップテーブルにおける各格子点データを所定の基準に基づいて複数分を合成して得られる合成データを新たなルックアップテーブルの格子点データとして格納しておく。合成データは、複数分を合成した分だけビット幅が広がる。
【０１６８】
データのビット幅を広げると、同メモリ容量であっても、データを格納するメモリのゲート規模（メモリサイズ）が小さくなる。よって、ビット幅の広い合成データとしてメモリ（データ格納部）に格納することで、チップサイズのより小さなメモリを使用することができ、メモリコストを低減することができる。
【図面の簡単な説明】
【図１】特定用途向け半導体内に構成される記憶手段の一例であるメモリの容量と、メモリが占める面積（以下ゲート（gate）規模という）の関係を説明する図である。
【図２】色データを変換する際に使用されるＬＵＴに格納するデータの構造（メモリデータ構造）の第１例を説明する図である。
【図３】第１例の合成手法を適用した第１実施形態のデータ変換装置の基本的な構成例を示す図である。
【図４】第１実施形態のデータ変換装置における色データの変換に関する処理手順の概要を示したフローチャートである。
【図５】第１実施形態のデータ変換装置におけるデータ構造の具体例を示す図である。
【図６】第１実施形態のデータ変換装置の具体的な構成を示すブロック図である。
【図７】演算処理部における単位立方体判定処理（図４のステップＳ１００〜Ｓ１０２）を説明する図である。
【図８】演算処理部における４面体判定処理（図４のステップＳ１０４〜Ｓ１０６）を説明する図である。
【図９】基点Ｐ２１と４面体判定部による判定結果に基づく４頂点の座標位置との関係を示す図である。
【図１０】演算処理部における出力値算出処理（図４のステップＳ１０８）の補間係数算出処理（Ｓ１０２ｃ）と補間演算処理（Ｓ１０２ｄ）を説明する図である。
【図１１】メモリデータ構造の第２例を説明する図である。
【図１２】第２例の合成手法を適用した第２実施形態のデータ変換装置の基本的な構成例を示す図である。
【図１３】メモリデータ構造の第３例を説明する図である。
【図１４】第３例の合成手法を適用した第３実施形態のデータ変換装置の基本的な構成例を示す図である。
【図１５】第３実施形態のデータ変換装置におけるデータ構造の具体例を示す図である。
【図１６】第３実施形態のデータ変換装置の具体的な構成を示すブロック図である。
【図１７】テトラ・ハイドラ補間を利用して、ＲＧＢ入力からＹＭＣＫ出力への色変換を行なう従来のデータ変換装置の構成例を示す図である。
【図１８】図１７に示すデータ変換装置におけるデータ格納部に使用される各頂点データの参照用メモリのデータ構成例を示す図である。
【符号の説明】
１…データ変換装置、２…データ格納部、４…演算処理部、６…アドレスバス、８…データバス、１０…アドレス生成部、１４…４面体判定部、１６…データ分離部、２０…補間係数生成部、２２…出力色データ決定部、２２Ｙ，２２Ｍ，２２Ｃ，２２Ｋ…補間処理部[0001]
TECHNICAL FIELD OF THE INVENTION
The present invention relates to a data conversion method and a data conversion device for generating output color data from input color data. More specifically, for example, an image processing apparatus such as a printer apparatus, a facsimile apparatus, or a multifunction peripheral having the functions thereof, for example, a multidimensional input of R (red), G (green), and B (blue) The image data of a predetermined dimension is converted to another color data such as converting the image data into one color data of each of multi-dimensional output image data of Y (yellow), M (magenta), C (cyan), and K (black). It relates to technology for converting data.
[0002]
[Prior art]
[Patent Document 1]
JP-A-3-229573
[Patent Document 2]
JP-A-8-181874
[Patent Document 3]
JP-B-58-16180
[Patent Document 4]
JP-A-56-14237
[0003]
2. Description of the Related Art Image forming apparatuses having a printing function, such as printer apparatuses and copying apparatuses, are used in various fields. Today, color image forming apparatuses are being used as various expression means for users. For example, a color page printer using an electrophotographic process (xerography) has attracted attention in terms of high quality image quality or high speed printing.
[0004]
Also, today, with the spread of DTP (DeskTop Publishing / Prepress), opportunities for handling color images are increasing. In this case, devices for inputting color images are mainly scanners, video cameras, and the like. The output devices are various color printers such as an ink jet system, a dye sublimation system, and electrophotography. Above all, attention has been paid to a printing function using an electrophotographic process capable of high-speed output.
[0005]
Each of these color input / output devices has a unique color space, and when color image data obtained from a certain scanner is directly transferred to another color printer and an image is output, the color of the image is changed. It is almost impossible to match the color of the original image. In order to make the two colors match, a color space conversion process is performed such that a color space of a so-called input device (such as a scanner or a video camera) is converted into a color space of an output device (the various color printers described above). Is required.
[0006]
Such color space conversion processing means that image signals of colors (in general, three colors of R (red), B (blue), and G (green)) obtained by an input device are converted into three colors on an output device side. Alternatively, the image signal is nonlinearly converted into image signals of four colors.
[0007]
Nonlinear conversion (for example, gamma conversion or log conversion) of digital signals such as images is generally performed using an LUT (Look Up Table). The method using an LUT is a method in which an output value corresponding to an input signal is calculated in advance, the value is held as a table, and an output value corresponding to the input value is obtained without calculation. By applying this method, a complicated arithmetic circuit can be eliminated.
[0008]
If the non-linear conversion is performed by an arithmetic circuit, the arithmetic circuit becomes complicated and the circuit scale increases. For example, when performing an arbitrary non-linear conversion process on one 8-bit video data using an LUT, With a memory having a capacity of 256 bytes, the processing can be realized and a complicated arithmetic circuit can be eliminated. Incidentally, since the above-described conversion is for converting one image signal into another image signal having another property, the LUT used there is called a one-dimensional LUT.
[0009]
For example, a hard copy output device such as a printer handles a color image signal. These output devices form images on paper or the like using color materials of colors such as Y, M, C, and K. Therefore, it has a color space composed of Y, M, C, and K. On the other hand, there are a scanner, a digital still camera, and the like that are assumed as color image input devices, each of which has a unique color space. These have various color spaces, and typically have a color space composed of R, G, and B. Therefore, when a hard copy is output by an output device such as a printer, it is necessary to convert these color spaces.
[0010]
However, if the above-described process of converting the three-color image data of the input device into one of a plurality of colors of the output device is performed using only the three-dimensional LUT, a large-capacity memory is required. For example, if the R, G, and B signals are 8-bit signals, and the output Y, M, C, and K signals are also 8-bit signals, there are 16,777,216 combinations of input values (2 ＾ 8 × 2 ＾ 8 × 2 ＾ 8 ways; “＾” indicates a power), and is used to convert the input of R, G, B into one of Y, M, C, K. 16 megabytes (Mbytes) of memory is required. Therefore, conversion to four colors of Y, M, C, and K requires a memory as large as 64 megabytes, which is four times 16 megabytes.
[0011]
That is, an LUT corresponding to input 24 bits and output 8 bits is required. In this case, the memory capacity per output color is 16 M (mega) bytes, and the above-mentioned memory is required for the number of colors of the output device. Therefore, the actual memory capacity is as large as 48 to 64 Mbytes. Therefore, this memory capacity has a large impact on the system from the viewpoint of cost and is not realistic.
[0012]
In order to solve such a problem, when an LUT is used in the color conversion processing, it is general to reduce the memory capacity of the LUT to be used by using an interpolation operation together. For example, a data conversion mechanism based on a combination of an LUT and an interpolation process has been proposed. This conversion mechanism reduces the capacity of the LUT in the conventional LUT system by configuring an output value corresponding to only a plurality of upper bits of an input signal as an LUT. If an output value corresponding to the input signal does not exist, an interpolation process is performed using data near the spatial position indicated by the input value to calculate an output value.
[0013]
Such interpolation calculation processing can be classified into several methods depending on how many pieces of data (grid point data) read from the LUT are used and on what kind of relation grid point data is used. In general, in a method using a lot of grid point data, the interpolation accuracy is improved, but the scale of the interpolation circuit is increased. As an interpolation method having a relatively small circuit scale, for example, a four-point interpolation method (Tetra-Hydra interpolation) described in Patent Documents 2, 2, and 3 is known. Patent Document 2 describes that there is also a five-point interpolation method. Patent Document 3 describes, as a conventional example, that a cubic interpolation method (8-point interpolation method) is also available as an interpolation method having a large circuit scale but high interpolation accuracy.
[0014]
The four-point interpolation method and the eight-point interpolation method described above use an interpolation method with the smallest circuit scale, that is, the least number of grid points, when performing an interpolation operation using grid point data existing corresponding to each grid point of the cubic grid. This corresponds to a method of interpolating with data and an interpolation method with the largest circuit scale, that is, a method of interpolating with the most grid point data. In other words, the number of pieces of grid point data (the number of pieces of data for interpolation calculation) at the time of performing an interpolation calculation using each piece of grid point data of the three-dimensional cubic grid is determined by the polyhedron internally dividing the unit cube determined by the grid points of the original LUT When the target is a cubic lattice, the number is a minimum of four and a maximum of eight.
[0015]
For example, a grid consisting of data of three input elements is formed, and which grid body contains the corresponding data is determined from the upper bits of the three input elements, and the grid is further divided into six tetrahedrons. A method of determining which tetrahedron contains the corresponding data from the lower bits of the three elements and calculating based on the basic data obtained from the vertices of the tetrahedron and the multiplier determined for each tetrahedron is a tetrad. Hydra interpolation calculation (tetrahedral interpolation calculation). In this case, if the input data is determined, the required tetrahedron is uniquely determined, and the tetrahedron may be determined only by a total of four vertices obtained from two line segments that do not intersect each other.
[0016]
On the other hand, in the interpolation processing, the interpolation accuracy of the output value differs depending on the number of neighboring data to be calculated. If more data is subjected to the interpolation calculation, the interpolation accuracy is improved. However, when a large amount of data is used, the scale of the interpolation circuit becomes large, and the effect of circuit reduction is reduced.
[0017]
As one method for solving this problem, Patent Literatures 1 and 2 propose methods for efficiently defining the amount of data held in an LUT and the number of data used for interpolation. For example, according to the method described in Patent Document 2, first, a lattice (unit cubic) composed of input three-element data is formed, and based on the upper bit data RH, GH, and BH of the three input elements, any lattice is determined. It is determined whether the corresponding data is included, and the data is further divided into six tetrahedrons. Based on the lower bit data RL, GL, and BL of the three input elements, which tetrahedron includes the corresponding data? Is determined, and an output value corresponding to the input value is calculated by a tetra-hydra method based on basic data obtained from the tetrahedron vertices and a multiplier (interpolation coefficient) determined for each tetrahedron. is there.
[0018]
In the method described in Patent Document 2, if input data is determined, a required tetrahedron is uniquely determined. That is, four neighboring grid points on the three-dimensional LUT of the inputs R, G, and B are specified. In order to determine the tetrahedron, it suffices to have a total of four vertices obtained from two line segments that do not intersect each other. These line segments can be represented by connecting successive vertices of one of the three input elements.
[0019]
Although the detailed description is omitted, when the method described in Patent Literature 2 is used, the number of data stored in the LUT is 17 data × when the input R, G, and B are converted into any one color (output color). When 17 data × 17 data (4913 grid points) is specified, the capacity of the LUT can be realized with 49 kilobytes, and a large capacity reduction can be achieved as compared with 16 megabytes in a system not using interpolation processing.
[0020]
[Problems to be solved by the invention]
However, in the method described in Patent Literature 2 or the like, when the time allocated as the processing time of the color space conversion for one pixel is extremely short, a plurality of data are read from the memory used as the LUT a plurality of times. Since it is impossible to do so, the same amount of memory must be used for the number of data for the interpolation operation site.
[0021]
In the case of the tetra-hydra interpolation, assuming the conversion from RGB to YMCK, four are used for extracting four grid points for conversion to Y, four for conversion to M, and four for conversion to C. And a total of 16 memories of the same capacity are required for conversion into K.
[0022]
FIG. 17 is a diagram illustrating a configuration example of a conventional data conversion device that performs color conversion from RGB input to YMCK output using tetra-hydra interpolation. FIG. 18 is a diagram illustrating an example of a data configuration of a reference memory of each vertex data used in the data storage unit 2 in the data conversion device 1 illustrated in FIG. As the memory, a RAM (Random Access Memory) configured in an application-specific semiconductor is used.
[0023]
In the device configuration shown in FIG. 17, four are used to extract four grid points for conversion to Y, four for conversion to M, four for conversion to C, and four for conversion to K. Memory is required. As shown in FIG. 18, each address of the reference memory (RAM) for converting the RGB input color data into the YMCK output color data includes the corresponding point data of the color conversion and the data bit width of the input three elements. It is stored by itself (8 bits in this example).
[0024]
Thus, in the conventional method, in order to obtain the data of the four vertices required for the tetra-hydra interpolation operation in a certain unit time, four memories having the same capacity and the same contents are prepared for one component of the output. There must be. In the case of this example in which the output is composed of four components Y, M, C and K, it is necessary to prepare a total of 16 memories.
[0025]
When the number of data to be stored in the LUT is defined as 17 data × 17 data × 17 data as described above, it is necessary to prepare 786 kilobytes as the total memory capacity instead of 49 kilobytes described above. Assuming importance on interpolation accuracy and assuming 8-point interpolation, 32 memories are required, and the total memory capacity is 1,568 kilobytes.
[0026]
Further, it is desired that a hard copy output device such as a printer performs a different color conversion for each component (for example, an image object such as a text, a computer graphics image, and a photographic image) in the output document. In this case, since a plurality of data is required as the data of each vertex, the number of memories used further increases, and naturally, the total memory capacity further increases.
[0027]
As described above, if the total memory capacity increases, the occupied area of the memory configured in the semiconductor increases, and a large chip size is required. As a result, the cost is increased.
[0028]
The present invention has been made in view of the above circumstances, and it is an object of the present invention to provide a data conversion method and apparatus capable of reducing an area occupied by a memory when performing color conversion using an interpolation operation. .
[0029]
[Means for Solving the Problems]
Although details will be described in the embodiments, the present invention has been made based on technical knowledge that, even when a memory having the same capacity is configured, the gate scale (memory size) is smaller when the bit width is increased. It is a thing.
[0030]
That is, the data conversion method and apparatus according to the present invention store the combined data obtained by combining a plurality of pieces of each grid point data in the original look-up table based on a predetermined reference as the grid point data of the new look-up table. Keep it. Then, at the time of data conversion processing, normal data conversion processing is performed after reading out the synthesized data and separating the data into original individual grid point data.
[0031]
For example, the first data conversion method according to the present invention uses the input color data for each pixel data of the input color data based on the grid point data of the lookup table used when converting the input color data into the output color data. A data conversion method for determining output color pixel data corresponding to data. First, each grid point data in an original lookup table for a plurality of output colors is added to each grid point of a lookup table in a predetermined manner. Synthesized data obtained by synthesizing a predetermined number of output colors based on the reference is stored.
[0032]
At the time of the data conversion processing, the composite data corresponding to the pixel data of the input color data is read from the data storage unit in which the composite data is stored, and the read composite data is separated into each grid point data for the output color. Then, for each output color, pixel data for the output color is determined using the separated grid point data for the output color.
[0033]
In the second data conversion method according to the present invention, first, each grid point in the original look-up table for a plurality of image attributes is stored in each grid point of the look-up table by a predetermined number of images based on a predetermined standard. The combined data obtained by combining the attributes is stored.
[0034]
At the time of the data conversion processing, the composite data corresponding to the pixel data of the input color data is read from the data storage unit in which the composite data is stored, and the read composite data is separated into each grid point data for the image attribute. Then, for each output color, pixel data for the output color is determined using the separated grid point data for the output color.
[0035]
A third data conversion method according to the present invention is a combination of the first and second data conversion methods. First, a plurality of image attributes and a plurality of output colors are assigned to each grid point of the lookup table. The composite data obtained by combining each grid point data in the original look-up table with a predetermined number of image attributes and a predetermined number of output colors based on a predetermined reference is stored.
[0036]
At the time of the data conversion process, the composite data corresponding to the pixel data of the input color data is read from the data storage unit in which the composite data is stored, and the read composite data is separated into each grid point data for the image attribute and the output color. . Then, for each output color, pixel data for the output color is determined using the separated grid point data for the output color.
[0037]
A first data conversion device according to the present invention is a device that implements the first data conversion method according to the present invention. First, each grid point data in an original lookup table for a plurality of output colors is It is provided with a data storage unit in which synthesized data obtained by synthesizing a predetermined number of output colors based on a predetermined standard is stored at each grid point.
[0038]
Further, the first data converter includes a combined data reading unit that reads combined data corresponding to the pixel data of the input color data from a data storage unit that stores the combined data, and a combined data reading unit that reads the combined data read by the combined data reading unit. A data separation unit that separates each output color into grid point data, and output color data that determines pixel data for the output color using the output color grid point data separated by the data separation unit for each output color And a decision unit.
[0039]
A second data conversion device according to the present invention is a device that performs the above-described second data conversion method according to the present invention. First, each grid point data in the original lookup table for a plurality of image attributes is It is provided with a data storage unit in which synthesized data obtained by synthesizing a predetermined number of image attributes based on a predetermined standard is stored at each grid point.
[0040]
The second data converter includes a combined data reading unit that reads combined data corresponding to the pixel data of the input color data from a data storage unit that stores the combined data, and a combined data reading unit that reads the combined data read by the combined data reading unit. A data separating unit that separates each image data into grid point data, and output color data that determines output color pixel data using the output color grid point data separated by the data separating unit for each output color And a decision unit.
[0041]
A third data conversion device according to the present invention is a device that implements the third data conversion method according to the present invention, and first, each of a plurality of image attributes and a plurality of output colors in an original lookup table for a plurality of output colors. A data storage unit is provided which stores, at each grid point, synthesized data obtained by synthesizing grid point data for a predetermined number of image attributes and a predetermined number of output colors based on a predetermined standard.
[0042]
Further, the third data converter includes a combined data reading unit that reads combined data corresponding to the pixel data of the input color data from a data storage unit that stores the combined data, and a combined data reading unit that reads the combined data read by the combined data reading unit. A data separation unit that separates the image attributes and the output color into respective grid point data, and pixel data for the output color are determined using the output color grid point data separated by the data separation unit for each output color. And an output color data determination unit.
[0043]
The dependent claims define further advantageous specific examples of the data conversion method and device according to the present invention. It should be noted that the combined data reading unit and the output color data determining unit can be realized not by hardware but by software, which is suitable for realizing the data conversion device according to the present invention by software using an electronic computer. A program and a computer-readable storage medium storing the program can also be extracted as the invention of the present application. The program may be provided by being stored in a computer-readable storage medium, or may be distributed via a wired or wireless communication unit.
[0044]
In addition, synthesized data obtained by synthesizing a plurality of pieces of each grid point data in the original lookup table based on a predetermined criterion is distributed via a wired or wireless communication means, and is stored in a data storage unit. It may be stored. Alternatively, the composite data may be provided stored in a computer-readable storage medium. In this case, before using the data conversion device, the synthesized data is read from the storage medium and stored in the data storage unit. Such a computer-readable storage medium storing the above-described composite data can also be extracted as the invention of the present application.
[0045]
[Action]
In the above configuration of the present invention, first, as a data storage unit for storing grid point data required for data conversion processing, first, a plurality of grid point data in the original look-up table are divided based on a predetermined reference. Prepare data in which the synthesized data obtained by synthesis is stored at each grid point. Then, the combined data is read from the data storage unit and separated into original individual grid point data, and then a normal data conversion process is performed.
[0046]
BEST MODE FOR CARRYING OUT THE INVENTION
Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.
[0047]
<< Basic structure of memory data >>
FIG. 1 is a diagram for explaining the relationship between the capacity of a memory, which is an example of a storage means included in an application-specific semiconductor, and the area occupied by the memory (hereinafter referred to as a gate scale). Here, FIG. 1A is a chart showing an example of a gate scale of a memory when a memory having the same capacity is configured under a plurality of conditions, and FIG. 1B shows the data of FIG. Is a graph. The data shown in FIG. 1A is quoted from “Toshiba Corporation TC260E ASIC Series Data Book”.
[0048]
FIG. 1A illustrates a case where a 1,048,576-bit memory is constructed under five conditions. It shows how the gate scale changes when configuring a memory of the same capacity while changing the bit width while using a common number of words. In FIG. 1B, the polygonal line indicates the total capacity in each configuration ((1) to (5)), and all have the same capacity (1,048,576 bits). On the other hand, the bar graph shows the gate scale in each configuration. Each configuration has the same number of words, but different bit width conditions.
[0049]
As can be seen from FIG. 1B, even when a memory having the same capacity (total capacity) is configured, the gate scale becomes smaller when the bit width is increased, as shown by configuration (5). When the gate size is reduced, the area occupied by the gate is reduced, resulting in a memory having a small chip size. As a result, even when a memory having the same capacity is used, by increasing the data bit width, a low-cost memory having a small chip size can be used.
[0050]
From this, in the data conversion device, if the data configuration is such that the bit width can be increased, a memory having a small chip size can be used as the memory configuring the LUT, thereby reducing costs. It is speculated that it can be done.
[0051]
<< First configuration example of memory data and corresponding circuit >>
FIG. 2 is a diagram illustrating a first example of a data structure (memory data structure) stored in an LUT used when converting color data. Based on the above consideration, when configuring memories of the same capacity, the gate size (memory size) is reduced by increasing the bit width. Here, as a technique for increasing the bit width, the first example focused on the following points.
[0052]
First, reference is made to the conventional data structure shown in FIG. As shown in FIG. 18, at each address of the reference memory (RAM), the corresponding point data of the color conversion is stored in the data bit width itself of the input three elements. Specifically, it is stored in 8 bits which is data of one component. On the other hand, in data conversion processing for performing color conversion from RGB input to YMCK output using tetra-hydra interpolation, reference data of Y, M, C, and K at the same address is used.
[0053]
Therefore, in the data structure of the first example, first, the number (the number of vertices) equal to the number of lattice points (vertexes) of the LUT required for the interpolation operation so that data required for the interpolation operation can be simultaneously extracted within a certain unit time. T (4 in the case of tetra-hydra interpolation). On the other hand, as shown in FIG. 2, some or all data of the output color components (Y, M, C, K in this example) are stored in the address positions of the respective lattice points of the newly configured LUT. They are stored together as composite data. Needless to say, the bit width is wider when all the pieces are combined as shown in FIG. 2A than when only a part is combined as shown in FIG. 2B.
[0054]
Here, the data bit width of an arbitrary lattice point to be synthesized is N bits, the data bit width of one component is M bits (M and N are positive integers, N> M), and the original (previous) The data size stored at an arbitrary address of the LUT is Y bits (Y is a positive integer), the number of output color components is X (X is a positive integer of 2 or more), and the output color component (X) is predetermined. Assuming that the data is divided into groups of number n (X = x1 + x2 +... + Xn; xi is a positive integer), the bit width N or Ni (unit is bits) of the composite data of the LUT lattice points of each group is expressed by the following equation. It can be shown in (1). If xi = X, it means that all are put together, and there is only one group. As can be seen from the previous example, usually, the data bit width M of one component is equal to the data size Y of the LUT.
(Equation 1)

[0055]
Also, as long as the expression (1) is satisfied, the data can be divided into a plurality of groups as shown in FIGS. 2B and 2C. In addition, as shown in FIG. 2C, depending on the way of grouping, a group can be stored with the data bit width of one component itself. In this case, since it is divided into a plurality of groups, in this case, it is assumed that the number X of output components is 3 or more. This is because the number X of output components must be a plurality (2 or more) in order to combine them, and when x = 2, only one group can be created by combining them.
[0056]
Note that the color data conversion processing is not limited to color conversion from a RGB input to a YMCK output using a multidimensional LUT. For example, a process using a one-dimensional LUT for converting data on a certain color axis (for example, each axis in a Lab space) into different data values on the same axis may be used. However, a one-dimensional LUT cannot convert color data of one input component into color data of a plurality of output components, and there is no concept of “combining a plurality of output colors”. The application range of the first example is limited to those in which the number of input components is “2” or more.
[0057]
Further, as an example in which the number of input components is “2”, for example, for an input component having three or more inputs, color conversion using a two-dimensional LUT focusing only on two axes is conceivable. More specifically, in a case where color data represented in the Lab color space is input, attention is paid only to a and b representing hue and saturation, and output values a and b are output using a two-dimensional LUT in ab space. Is conceivable. This means that the input value is maintained for the brightness, but the data value is changed (color conversion) for the hue and saturation. When the above-described combining method of the first example is applied to this case, the output values a and b may be combined.
[0058]
As an example in which the number of input components is "3", color space data such as RGB, sRGB, Lab, YCbCr, YMC, and XYZ can be considered. Further, as an example in which the number of input components is “4”, color space data such as YMCK, K (black amount) + Lab, RGB + L (lightness) can be considered. Further, the input color space and the output color space are not limited to the same one as in the color conversion from the RGB input to the YMCK output, but may be different.
[0059]
As described above, the data bit width is increased by combining the data of the same address of the original LUT grid point data for a plurality of output color components by applying the data synthesis method of the first example. If the LUT data is stored in the memory in a state of large size, the gate scale (memory size) can be reduced, and the memory cost can be reduced. This makes it possible to provide an efficient LUT configuration method and an access method with low redundancy.
[0060]
FIG. 3 is a diagram illustrating a basic configuration example of the data conversion device 1 according to the first embodiment when an LUT having a data structure to which the synthesis method of the first example illustrated in FIG. 2 is applied is used.
[0061]
The data conversion apparatus 1 of the first embodiment receives an input digital color separation signal of Y1 bits of each of the input color components I1, I2,..., Iw from an external device. The data conversion device 1 performs color conversion of input data into Y2-bit data of output color components S1, S2,..., Sx. For example, when 8-bit data of each of three colors R, G, and B is input, color conversion is performed to 8-bit data of each of four colors of Y, M, C, and K. Note that the format of the input data and output data is merely an example, and other color series may be used, and the number of bits Y1 and Y2 are not limited to “8”.
[0062]
As shown in the drawing, a data conversion device 1 according to the first embodiment includes a data storage unit 2 that stores basic data used when obtaining output data after color conversion corresponding to input data by interpolation calculation, An arithmetic processing unit for obtaining output pixel data corresponding to input pixel data by interpolation for each output color while exchanging data with the unit;
[0063]
The data storage unit 2 stores the basic data corresponding to the grid point data associated with each grid point in the w-dimensional grid space (w-dimensional LUT) defined by the input color signal elements as described with reference to FIG. , A plurality of output color components are grouped according to a predetermined standard, and a plurality of basic data at the same address in the group are collected and stored as one composite data, thereby forming a lookup table LUT. It has become. For example, when all the output colors are grouped into one group, reference memories 2T1, 2T2,..., 2Tt are used for the number T of grid point data required for the interpolation operation. Each of the reference memories 2T1, 2T2,..., 2Tt stores the combined grid point data for the output colors S1 to Sx.
[0064]
Each of the memories 2T1, 2T2,..., 2Tt is independently addressable so as to be able to cope with the case where the time allocated as the processing time of the color space conversion for one pixel is very short. It is configured to be readable, that is, independently readable.
[0065]
The arithmetic processing unit 4 includes, as signal processing systems, S1 signal generation units 4S1, S2 signal generation units 4S2,..., Sx signal generation units 4Sx so as to correspond to the color components S1, S2,. including. Further, the arithmetic processing unit 4 includes a data separation unit 16 that separates the combined data stored in the data storage unit 2 into data for each output component. Here, in the data separation unit 16 of the first embodiment, an individual circuit is provided inside corresponding to the output colors S1, S2,..., Sx. The arithmetic processing unit 4 may be composed of, for example, a CPU (Central Processing Unit) or a DSP (Digital Signal Processor).
[0066]
An address bus 6 for sending a control signal such as an address signal so that the arithmetic processing unit 4 reads basic data from the data storage unit 2 and a data read from the data storage unit 2 between the data storage unit 2 and the arithmetic processing unit 4. The data bus 8 sends the basic data to the arithmetic processing unit 4.
[0067]
The arithmetic processing unit 4 is provided with an address generation unit that generates an access address to a look-up table in which data used for an interpolation operation for generating an output value is stored. The address bus 6 transmits an address signal, a clock, a chip select signal, or the like for setting a basic data read position in each memory space (memory 2T1, 2T2,..., 2Tt) in the data storage unit 2. Be
The data bus 8 is used to input input data before data conversion from an external device (not shown) or output output data after conversion to the same or another external device.
[0068]
In such a configuration, the arithmetic processing unit 4 first obtains a plurality of output color components from the grid point data of each output color component from an arbitrary address in the memories 2T1, 2T2,. The N-bit or Ni-bit synthesized data synthesized for the element is read. The data separation unit 16 separates the read composite data into grid point data for each output color component. The signal generators 4S1 to 4Sx provided for the respective color components S1, S2,..., Sx perform complementary operations using the separated grid point data, thereby outputting the respective color components S1, S2,. Generate data.
[0069]
FIG. 4 is a flowchart illustrating an outline of a processing procedure regarding color data conversion in the data conversion device 1 according to the first embodiment. In the following description, it is assumed that the input data is 8 bits R, G, B and the output data is 8 bits Y, M, C, K, and that the arithmetic processing unit 4 performs a tetra-hydra interpolation operation. In this case, as shown in FIG. 3, the arithmetic processing unit 4 uses a Y signal generation unit 4Y, an M signal generation unit 4M, a C signal generation unit 4C, and a K signal generation unit 4K. The memory 2 uses four memories 2T1, 2T2, 2T3, and 2T4.
[0070]
The arithmetic processing unit 4 of the first embodiment uses the grid point data of the color conversion LUT (look-up table) stored in the data storage unit 2 to output each signal (R, G) input from an external device. , B) are converted into color data of output colors (Y, M, C, K) using an interpolation process using a tetra-hydra method.
[0071]
Therefore, when the image data R, G, and B are input, the arithmetic processing unit 4 first converts the input 8-bit R, G, and B digital color separation signals into upper α bits (most significant bits) for each pixel. It is divided into α consecutive bits from the bit side and lower β bits (β consecutive from the least significant bit side) (S100).
[0072]
Next, the arithmetic processing unit 4 configures a three-dimensional lattice (three-dimensional LUT) composed of R, G, and B data of the three input elements based on the upper α-bit data RH, GH, and BH of the three input elements. Then, it is determined which of the unit cubes (unit cubes in this example) defined by each grid point contains the corresponding data (S102). The processing of steps S100 to S102 is called a unit cube determination processing.
[0073]
Next, the arithmetic processing unit 4 further divides the corresponding unit cube into six tetrahedrons (S104), and based on the data RL, GL, BL of the lower β bits of the input three elements, It is determined whether the corresponding data is included (S106). The processing of steps S104 to S106 is referred to as tetrahedron determination processing.
[0074]
Next, the arithmetic processing unit 4 obtains an output value (pixel data to be output) corresponding to the input value (input pixel data) by the tetra-hydra method of interpolation calculation in the corresponding tetrahedron (S108). ).
[0075]
More specifically, as shown on the right side of the figure, the arithmetic processing unit 4 first specifies four grid points required for the interpolation operation in the data group of the same address based on the coordinate values of the base point including the input point. Then, a four-grid point specifying process (also referred to as an access address specifying process) is performed (S108a). That is, the arithmetic processing unit 4 assigns an access address to the LUT in which the data used to generate the output value by the interpolation arithmetic processing by the tetra-hydra method is stored in the four memories 2T1, 2T2, 2T3, and 2T4. Identify.
[0076]
At an arbitrary address of the LUT, data for the number of output components is synthesized and stored in n bits (32 bits in this example). Therefore, the data separating unit 16 of the arithmetic processing unit 4 extracts grid point data from the specific address corresponding to the input value specified in the access address specifying process, and outputs the extracted 32-bit grid point data. It is separated into m-bit (8-bit in this example) data which is unit data of a color component (S108b).
[0077]
In parallel with the four-lattice-point specifying process (S108a), the arithmetic processing unit 4 assigns a weighting factor to the four basic data (in this example, the original grid point data itself) obtained from the tetrahedral vertices for the corresponding tetrahedron. An interpolation coefficient calculation process of calculating an interpolation coefficient that is a multiplication coefficient) is performed (S108c). Then, the arithmetic processing unit 4 calculates the corresponding tetrahedron based on the basic data (grid point data) obtained by the four grid point specifying process (S108b) and the interpolation coefficient obtained by the interpolation coefficient calculation process (S108c). Then, an interpolation operation for obtaining an output value Dout corresponding to the input value is performed (S108d).
[0078]
The arithmetic processing unit 4 performs the above-described series of processing for each of the output colors Y, M, C, and K (S112), and performs the processing for all the pixels (S110) to obtain the color of the input RGB color space. The data is converted to data in the YMCK color space.
[0079]
FIG. 5 is a diagram illustrating a specific example of a data structure when performing color conversion from RGB input to YMCK output using a multidimensional LUT in the data conversion device 1 of the first embodiment. For comparison with the conventional example shown in FIG. 18, the number of data stored in the original (conventional) LUT is defined as 17 data × 17 data × 17 data (4913 grid point addresses are required).
[0080]
Here, as described above, the arithmetic processing unit 4 performs the tetra-hydra interpolation operation. In this case, the memory (data storage unit 2) includes four memories (RAM0 to RAM4) for four vertices so that data of four vertices necessary for the tetra-hydra interpolation operation can be simultaneously extracted within a certain unit time. RAM 3).
[0081]
At the address position of each vertex, grid point data for four colors of YMCK are collected and held as one combined data. Here, the RGB input data is 8 bits, and four colors of YMCK are combined, so that one word is stored in a state of 8 bits × 4 colors = 32 bits. According to such a configuration, the occupied area of the memory configured in the application-specific semiconductor is significantly smaller than the case where the memory is individually used for each output color, and the cost is economical. Is overwhelming.
[0082]
FIG. 6 is a block diagram illustrating a specific configuration of the data conversion device 1 according to the first embodiment. The configuration of the data converter 1 shown in FIG. 6 is a specific circuit configuration to which the memory configuration shown in FIG. 5 is applied in the case of three-dimensional input of R, G, and B. Here, it is also assumed that the components of the input digital color separation signal are R, G, and B, each of which is 8-bit data.
[0083]
As illustrated, the arithmetic processing unit 4 includes an address generation unit 10 that executes a unit cube determination process (Steps S100 to S102 in FIG. 4) and an access address identification process (S108a), and a tetrahedron determination process (S104 to S108). The tetrahedron determination unit 14 that performs S106), the data separation unit 16 that performs data separation processing (S108b), the interpolation coefficient generation unit 20 that performs interpolation coefficient calculation processing (S108c), and the four neighboring data An output color data determination unit 22 for executing a tetra-hydra interpolation calculation process (S108d) to be used.
[0084]
The data storage unit 2 refers to a first vertex data reference memory 2T1 and a second vertex data that constitute an LUT for referencing the data of the first vertex so as to correspond to the four vertices in the tetra-hydra interpolation operation. A second vertex data reference memory 2T2 which forms an LUT for performing the operation, a third vertex data reference memory 2T3 which forms an LUT for referring to the third vertex data, and a fourth vertex data which is referred to. There is provided a fourth vertex data reference memory 2T4 constituting the LUT.
[0085]
The memories 2T1 to 2T4 store LUTs having the same contents and the same capacity. Conventionally, when performing tetra-hydra interpolation calculation, as shown in FIG. 18, four memories having the same capacity and the same contents are prepared, and a total of 16 memories are prepared in accordance with four outputs of Y, M, C, and K. Had to be prepared. Here, although the address to each vertex is different, each color requires data of the same vertex, and the address value is the same even if the actual memory to be accessed is different.
[0086]
Therefore, as shown in FIGS. 3A and 5, by combining the four output value components of one address, it is possible to simplify the memory. In this case, the output value is 8-bit data, and the output color components are four colors of Y, M, C, and K, so that one address stores 32-bit composite data.
[0087]
For example, in the first vertex data reference memory 2T1, composite data "Y1 + M1 + C1 + K1" in which data on the first vertices of the four output colors Y, M, C, and K are combined (the reference numeral "1" indicates the vertex number). Is stored. Similarly, the composite data “Y2 + M2 + C2 + K2” is stored in the second vertex data reference memory 2T2, the composite data “Y3 + M3 + C3 + K3” is stored in the third vertex data reference memory 2T3, and the composite data “T3” is stored in the fourth vertex data reference memory 2T4. Y4 + M4 + C4 + K4 ″ are stored.
[0088]
With such a memory configuration, the advantage of minimizing the gate scale can be enjoyed. As described above, when configuring memories of the same capacity, it is possible to reduce the gate size by increasing the bit width and decreasing the number of memories.
[0089]
The data separation unit 16 of the first embodiment separates combined data for all output colors into output color components, and outputs the first output data so as to correspond to the four vertex memories 2T1 to 2T4. A color data separation unit 16T1, a second output color data separation unit 16T2, a third output color data separation unit 16T3, and a fourth output color data separation unit 16T4 are provided.
[0090]
The output color data determination unit 22 includes a Y interpolation processing unit 22Y, an M interpolation processing unit 22M, a C interpolation processing unit 22C, and a K interpolation processing unit 22K corresponding to the four output colors Y, M, C, and K. Have.
[0091]
The address generating unit 10 is an example of a combined data reading unit, and has both functions of a unit cell determining unit and an address setting unit. Further, the interpolation coefficient generation unit 20 and the output color data determination unit 22 constitute an interpolation calculation processing unit.
[0092]
FIG. 7 is a diagram for explaining the unit cube determination processing (steps S100 to S102 in FIG. 2) in the arithmetic processing unit 4, and includes an original three-dimensional LUT constituting grid point data, which is a representative value of an output value, and an input. FIG. 3 is a diagram illustrating a relationship with data. Here, FIG. 7A is a conceptual diagram of the LUT in a three-dimensional space, and FIG. 7B is a three-dimensional LUT in which each 8-bit RGB input signal corresponds to the upper 4-bit signal of each signal. The above positional relationship is shown.
[0093]
As shown in FIG. 7A, the range of possible values for each of the R, G, and B axes is divided into a plurality. Here, an example in which each axis is divided into 16 is shown. An LUT is configured by using a grid point obtained by such division as an address and associating the converted value (representative value) at that grid point. Therefore, in this example, it is possible to create a three-dimensional LUT in which, for 17 × 17 × 17 = 4913 grid points, corresponding converted data is stored as a representative value.
[0094]
The arithmetic processing unit 4 converts the inputted 8-bit R, G, B digital color separation signal into upper α bits (α consecutive from the most significant bit) and lower β bits (β consecutive from the least significant bit). ). For example, the arithmetic processing unit 4 sets α = β = 4 and divides the input R, G, and B into upper four bits RH, GH, and BH and lower four bits RL, GL, and BL, respectively. Of course, α and β may be other than these. Note that “α + β” is preferably equal to the number of bits 8 of the input data.
[0095]
For example, as shown in FIG. 7A, each LUT holds 17 data for each axis of RGB. Here, as shown in FIG. 7B, in the unit cube formed by the combination value of the upper 4-bit data, the combination value of the arbitrary input R, G, B 8-bit data (input pixel data) Since the spatial position P22 is included, a unit cube (unit lattice) starting from the spatial position P21 indicated by the upper 4-bit data RH, GH, and BH is specified as a unit cube that includes the input 8-bit data. Becomes possible.
[0096]
Each vertex constituting the unit cube can be generated based on the spatial position P21. That is, with P21 (RH, GH, BH) as a base point, P23 (RH + 1, GH, BH), P24 (RH + 1, GH + 1, BH), P25 (RH, GH + 1, BH), P26 (RH, GH, BH + 1), P27 (RH + 1, GH, BH + 1), P28 (RH, GH + 1, BH + 1), and P29 (RH + 1, GH + 1, BH + 1) are the coordinates of the vertices of the unit cube. In the case of the present embodiment, the coordinates of these vertices match the coordinates of the lattice points of the original three-dimensional LUT. Further, the calculation formula of the access address to the LUT is the following formula (2).
(Equation 2)

[0097]
The arithmetic processing unit 4 calculates the output value corresponding to the combination value of the 8-bit data of the inputs R, G, and B using the four grid point data near the input point, that is, the unit shown in FIG. It is obtained by the Tetra Hydra method using four of the coordinate data of the eight vertices of the cube as basic data.
[0098]
FIG. 8 is a diagram for explaining the tetrahedron determination process (steps S104 to S106 in FIG. 4) in the arithmetic processing unit 4, and includes a unit cube that includes input data and a unit cube divided into six. It is a figure which shows the relationship with the obtained tetrahedron. As described above, the arithmetic processing unit 4 of the first embodiment stores each of the output signals YMCK corresponding to the combination value of the input digital color separation signals R, G, and B in the data storage unit 2. Based on the basic data (lattice point data) present, it is obtained by an interpolation operation by the tetra-hydra method.
[0099]
The unit cube shown in FIG. 7B can be further divided into six unit tetrahedrons having the same volume, as shown in FIG. The tetrahedron included in the spatial position P22 of the inputs R, G, and B can be specified as shown in Table 2 shown in the lower part of FIG. 8 based on the magnitude relation of the lower bit data RL, GL, and BL.
[0100]
For example, when the lower bit data RL, GL, and BL have a relationship of RL ≧ GL ≧ BL, inputs R, G, and B are (RH, GH, BH), (RH + 1, GH, BH), (RH + 1, GH + 1, BH) and (RH + 1, GH + 1, BH + 1). When RL>BL> GL, the inputs R, G, and B are (RH, GH, BH), (RH + 1, GH, BH), (RH + 1, GH, BH + 1), (RH + 1, GH + 1, BH + 1). It belongs to the tetrahedron 2 as the vertex. When BL ≧ RL> GL, the inputs R, G, and B are (RH, GH, BH), (RH + 1, GH, BH + 1), (RH, GH, BH + 1), (RH + 1, GH + 1, BH + 1). Belongs to a tetrahedron 3 having a vertex as a vertex.
[0101]
Similarly, when BL> GL ≧ RL, the inputs R, G, and B are (RH, GH, BH), (RH, GH, BH + 1), (RH, GH + 1, BH + 1), (RH + 1, GH + 1). , BH + 1) as vertices. When GL ≧ BL> RL, the inputs R, G, and B are (RH, GH, BH), (RH, GH + 1, BH), (RH, GH + 1, BH + 1), (RH + 1, GH + 1, BH + 1). Belong to a tetrahedron 5 having a vertex as a vertex. When GL> RL ≧ BL, the inputs R, G, and B are (RH, GH, BH), (RH, GH + 1, BH), (RH + 1, GH + 1, BH), (RH + 1, GH + 1, BH + 1). Belongs to a tetrahedron 6 having a vertex as a vertex.
[0102]
Therefore, first, when the image data R, G, and B are input, the address generation unit 10 of the arithmetic processing unit 4 converts the input data R, G, and B into upper bit data RH, GH, and BH (in this example, the upper 4 bits). Bits; collectively referred to as DH) and lower bit data RL, GL, BL (in this example, lower 4 bits; collectively referred to as DL), and based on the upper 4 bits of data DH in FIG. As shown in (B), the input position P22 of the unit cube including the input point in the three-dimensional LUT is specified. In addition, the address generation unit 10 inputs the lower bit data DL of the divided data to the tetrahedron determination unit 14 and the interpolation coefficient generation unit 20.
[0103]
Upon receiving the lower 4-bit signals DL of the input RGB signals generated by the address generator 10, the tetrahedron determination unit 14 compares the magnitudes of the three signals and calculates the input 8-bit signals. It is determined which tetrahedron in the unit cube is included. The tetrahedron determination unit 14 notifies the determination result to the address generation unit 10 and the interpolation coefficient generation unit 20.
[0104]
Assuming that the lower four-bit signal DL of the input RGB signal is RL, GL, and BL, respectively, assuming that the comparison condition in the tetrahedron determination unit 14 conforms to the tetrahedron configuration of FIG. Is the tetrahedron 1, the tetrahedron 2 when RL>BL> GL, the tetrahedron 3 when BL ≧ RL> GL, the tetrahedron 4 when BL> GL ≧ RL, and the GL ≧ BL> RL Is a tetrahedron 5, and a tetrahedron 6 when GL> RL ≧ BL.
[0105]
FIG. 9 is a diagram illustrating a relationship between the base point P21 and the coordinate positions of the four vertices based on the determination result by the tetrahedron determination unit 14. In the figure, a corresponds to the base point P21 shown in FIG. 7B.
[0106]
The input point P22 indicating the input data position is included in the unit cube defined by the upper bit data RH, GH, and BH of the input data, and the position of this unit cube is the base point P21 in FIG. (A).
[0107]
The address generator 10 receives the result of the tetrahedron determination process by the tetrahedron determiner 14 and determines the eight vertices (a to h in FIG. 9) of the unit cube including the input point P22 by the tetrahedron determiner 14. The four vertex addresses based on the result are calculated according to equation (2), and are set in the four memories 2T1, 2T2, 2T3, and 2T4. By doing so, when data is read from each of the memories 2T1, 2T2, 2T3, and 2T4 of the data storage unit 2, 32-bit composite data indicating grid point data of four vertices near the input point P22 is obtained from the set address. The data is input to the data separation unit 16.
[0108]
From the memories 2T1 to 2T4, data of the first to fourth vertices are output simultaneously within a certain unit time. The output data is composite data in which data of four identical vertices (same addresses) of Y, M, C, and K are integrated. For this reason, each of the output color data separation units 16T1 to 16T4 of the data separation unit 16 refers to a predetermined data synthesis method, and outputs unit data of output colors Y, M, C, and K, that is, an 8-bit unit. Separate into data. The data separated into 8-bit units is referred to as interpolation target data D1, D2, D3, and D4. Since the composite data is prepared in a format summarized for each output color, the data separation process is simple.
[0109]
As can be seen from Table 3 in FIG. 9, the lattice points assigned to the interpolation target data D1, D2, D3, and D4 are different for each of the six tetrahedrons. The data separation unit 16 inputs the interpolation target data D1, D2, D3, and D4 for each separated output color to the corresponding interpolation processing units 22Y to 22K.
[0110]
FIG. 10 is a diagram for explaining the interpolation coefficient calculation processing (S102c) and the interpolation calculation processing (S102d) of the output value calculation processing (step S108 in FIG. 4) in the calculation processing unit 4. Here, FIG. 10A is a chart showing a calculation formula of an interpolation coefficient required for the interpolation calculation, and FIG. 10B is a functional block diagram of the interpolation calculation processing.
[0111]
The interpolation coefficient calculation process (S102c) is executed by the interpolation coefficient generation unit 20. The interpolation coefficient generation unit 20 performs interpolation based on the lower 4-bit signals DL (RL, GL, BL) of the input RGB signals input from the address generation unit 10 and the tetrahedron determination result determined by the tetrahedron determination unit 14. The calculation coefficients W1, W2, W3, and W4 are calculated. Table 4 shown in FIG. 10A shows an equation for calculating an interpolation coefficient required for the interpolation operation by the tetra-hydra method. As can be seen from Table 4, the calculation formula is different for each of the six tetrahedrons.
[0112]
Upon receiving the result of the tetrahedron determination processing (steps S104 to S106 in FIG. 4), the interpolation coefficient generation unit 20 switches the calculation formula defined for each tetrahedron including the input value based on Table 3. Then, the interpolation coefficients W1, W2, W3, and W4 are calculated based on the lower bit data RL, GL, and BL of the four grid points forming the tetrahedron, and the interpolation coefficients W1, W2, W3, and W4 are calculated by the interpolation processing units 22Y to 22Y. Input commonly to 22K. This interpolation coefficient calculation processing is the same as the processing in the conventional data conversion device.
[0113]
The interpolation calculation processing (S102d) is executed by the output color data determination unit 22. Each of the interpolation processing units 22Y to 22K includes the interpolation coefficients W1, W2, W3, and W4 input from the interpolation coefficient generation unit 20 and the interpolation target data D1, D2, and D3 input from the corresponding data separation units 16T1 to 16T4. Based on D4 (basic data of four vertices), an interpolation calculation process by a tetra-hydra method of calculating a final output value Dout is performed. For example, the final output value Dout is calculated according to equation (3). FIG. 10B functionally shows the equation (3). In equation (3), the output value is obtained by linear interpolation. However, the present invention is not limited to this, and a quadratic equation or another higher-order interpolation calculation equation may be used.
[Equation 3]

[0114]
FIG. 11 is a diagram illustrating a second example of the memory data structure. The data structure of the first example collectively stores (combines) some or all of the output color components. On the other hand, the data structure of the second example is characterized in that some or all data of a plurality of image attributes are collectively (combined) and stored. Hereinafter, only differences from the data structure of the first example will be described.
[0115]
In the data structure of the first example, there is one neighboring data in the LUT corresponding to the input value at each vertex, and the data output by interpolation is determined one-to-one with respect to one input. Was. However, it is desired that a hard copy output device such as a printer performs a different color conversion for each component (eg, text, computer graphics image, photographic image, etc.) in the output document. That is, a plurality of data corresponding to different image attributes are required as data of each vertex.
[0116]
In this case, in the conventional configuration, the required number of memories for one image attribute (the number of grid points (vertexes) of the LUT required for the interpolation operation × the number of output color components X) is determined by the number of required image attributes. You have to prepare for the minute. For example, in order to convert RGB input color data into YMCK output color data by tetra-hydra interpolation, four conversions to Y are performed to extract four grid points, and similarly four conversions to M are performed. Since a total of 16 are required for conversion into C and four for conversion to K, if the number of image attributes is G, 16 · G memories are required, and the scale of the memory circuit is required. Increase.
[0117]
Therefore, in the data structure of the second example, the number of grid points (number of vertices) of the LUT required for the interpolation operation, which is the number of memories required for one image attribute, and the number of output color components are the same as in the conventional example. It is composed of the same number of memories as X. When converting RGB input color data into YMCK output color data by tetra-hydra interpolation, 4 × 4 = 16 memories are used.
[0118]
On the other hand, as shown in FIG. 11, which shows an example corresponding to four image attributes, the output position of each of the output color components is text, computer graphics image, or A part or all of the selected data components according to the image attribute such as a photographic image are collected (combined) and stored. It goes without saying that the bit width is wider when all the pieces are combined as shown in FIG. 11A than when only a part is combined as shown in FIG. 11B.
[0119]
Here, the data bit width of an arbitrary lattice point to be synthesized is L bits, the data bit width of one component is M bits (L and M are positive integers), and an arbitrary address of the original (previous) LUT. The stored data size is Y bits (Y is a positive integer), the number of selected data components according to the image attribute is Z (Z is a positive integer of 2 or more), and the number of selected data components (Z) is predetermined. If the data is divided into groups of number n (Z = z1 + z2 +... + Zn; zj is a positive integer), the data bit width L or Lj (unit is bit) of the LUT lattice point of each group is expressed by the following equation (4). ).
(Equation 4)

[0120]
If Zj = Z, it means that all are put together, and there is only one group. As can be seen from the previous example, usually, the data bit width M of one component is equal to the data size Y of the LUT.
[0121]
In addition, as long as the expression (4) is satisfied, as shown in FIG. 11B and FIG. Also, as shown in FIG. 11C, depending on the way of grouping, a group can be stored with the data bit width of one component itself.
[0122]
As shown in FIGS. 11A to 11C, there are various methods of data synthesis for integrating a plurality of image attributes. These methods are the same for each output color component. Preferably, a technique is employed. However, this does not exclude changing the method of data synthesis according to the output color component. For example, the method shown in FIG. 11A is used for the output colors Y, M, and C, while the method shown in FIG. 11B is used for the output color K.
[0123]
In addition, since the data is divided into a plurality of groups, in this case, it is assumed that the number Z of the selected data components is 3 or more. This is because the number Z of selected data components must be plural (2 or more) in order to combine them, and when Z = 2, only one group can be created by combining them.
[0124]
The second example is different from the first example in that a process using a one-dimensional LUT for converting data on a certain color axis (for example, each axis in a Lab space) to a different data value on the same axis, that is, an input component The number 1 is also applicable. This is because the conversion characteristics of the one-dimensional LUT may be switched according to the image attribute.
[0125]
An example in which the number of input components is "2", "3", or "4" is the same as in the first example. Since the specific example of the color space is the same as that of the first example, the description is omitted.
[0126]
FIG. 12 is a diagram illustrating a basic configuration example of the data conversion device 1 according to the second embodiment when an LUT having a data structure to which the synthesis method of the second example illustrated in FIG. 11 is applied is used. The difference from the configuration of the first embodiment shown in FIG. 3 is the memory data structure in the data storage unit 2 and the function of the data separation unit 16. Hereinafter, differences will be described.
[0127]
As described with reference to FIG. 11, the data storage unit 2 according to the second embodiment groups selected data components corresponding to a plurality of image attributes according to a predetermined criterion, and stores a plurality of bases of the same address in the group. The look-up table LUT is configured by storing the data in one piece of combined data. For example, when all the selected data components are put together into one group, reference is made to the number T × X, which is equal to the number of lattice points (vertexes) T × the number X of output color components of the LUT required for the interpolation operation. , 2S1T2,..., 2S1Tt, 2S2T1, 2S2T2,..., 2S2Tt, 2SxT1, 2S1T2,. For example, when RGB input color data is converted into YMCK output color data by tetra-hydra interpolation, 16 memories are used because T = 4 and X = 4. Each of the reference memories 2T1, 2T2,..., 2Tt stores the combined grid point data for the image attributes Z1 to Zz.
[0128]
Each of the memories 2S1T1, 2S1T2,..., 2SxTt is independently addressable so that it can cope with a case where the time allocated as the processing time of the color space conversion for one pixel is very short. It is configured to be readable, that is, independently readable.
[0129]
The arithmetic processing unit 4 includes, as signal processing systems, S1 signal generation units 4S1, S2 signal generation units 4S2,..., Sx signal generation units 4Sx so as to correspond to the color components S1, S2,. including. The arithmetic processing unit 4 separates the combined data stored in the data storage unit 2 into data for each image attribute (selected data component) such as a text, a computer graphics image, or a photographic image. A section 16 is provided. Here, the data separation unit 16 of the second embodiment has an internal circuit corresponding to grid points T1, T2,..., Tt of the LUT and output colors S1, S2,. Is provided.
[0130]
The arithmetic processing unit 4 firstly outputs the grid of each output color component from any address of the memories 2S1T1, 2S1T2,..., 2S1Tt, 2S2T1, 2S2T2,..., 2S2Tt, 2SxT1, 2S1T2,. L-bit or Lj-bit synthesized data which is point data and synthesized for a plurality of image attributes is read. The data separation unit 16 refers to a predetermined data synthesis method and converts the read synthesized data into grid point data for each image attribute such as a text, a computer graphics image, or a photographic image. Separate and select grid point data according to the image attribute to be processed. Since the composite data is prepared in a format summarized for each image attribute, data separation / selection processing is simple. The signal generators 4S1 to 4Sx provided for each of the color components S1, S2,..., Sx perform a complementary operation using the separated and selected grid point data, so that each of the color components S1, S2,. To generate output data.
[0131]
Note that the processing procedure regarding color data conversion in the data conversion device 1 of the second embodiment is the same as that of the first embodiment except that the data separation processing in the data separation unit 16 is separation / selection processing according to image attributes. This is the same as the processing procedure of the embodiment. The flowchart showing the processing procedure is not shown.
[0132]
According to the configuration of the second embodiment, the addresses of the vertices of the memories 2S1T1, 2S1T2,..., 2S1Tt, 2S2T1, 2S2T2,..., 2S2Tt, 2SxT1, 2S1T2,. At the position, grid point data of a plurality of image attributes are collected and held as one composite data. That is, in order to perform different color conversion for each component (image object) in the output document, the number of memories used (the number of grid points of the LUT (the number of vertices required for interpolation calculation) ) While maintaining T × the number of output color components X), the overall capacity is maintained by increasing the data bit width. According to such a configuration, even when the total memory capacity is increased to perform different color conversion for each image object in the output document, compared to the case where the memory is individually used for each image object, The memory occupied by the memory configured in the application-specific semiconductor is remarkably reduced, and the cost is economically significant.
[0133]
<< Third Configuration Example of Memory Data and Corresponding Circuit >>
FIG. 13 is a diagram illustrating a third example of the memory data structure. This third example is a combination of the first and second examples.
[0134]
First, similarly to the first example, the data structure of the third example is such that the number of grid points (number of vertices) of the LUT required for the interpolation operation can be set so that the data required for the interpolation operation can be simultaneously extracted within a certain unit time. (T; four in the case of tetra-hydra interpolation).
[0135]
On the other hand, as shown in FIG. 13 showing an example corresponding to four output colors and three image attributes, output color components (Y, M, C or K), and collectively (composite) and store some or all data of a plurality of image attributes. As shown in FIG. 13A, the bit width is wider when all of the plurality of output color components and the plurality of image attributes are combined than when only a part is combined as shown in FIG. 13B. Needless to say.
[0136]
Here, the data bit width of an arbitrary lattice point to be synthesized is K bits, the data bit width of one component is M bits (L and M are positive integers), and an arbitrary address of the original (previous) LUT. The stored data size is Y bits (Y is a positive integer), the number of output color components is X (X is a positive integer of 2 or more), and the output color components (X) are grouped into a predetermined number n. Dividing (X = x1 + x2 +... + Xn; xi is a positive integer), the number of selected data components according to the image attribute is Z (Z is a positive integer of 2 or more), and the number of selected data components (Z) is predetermined. If the data is divided into groups of number n (Z = z1 + z2 +... + Zn; zj is a positive integer), the data bit widths K, Ki, Kj, and Kij (units are bits) of the LUT lattice points of each group are as follows: Equation (5) can be used.
(Equation 5)

[0137]
If xi = X and Zj = Z, it means that all are put together, and there is only one group. When Xi = X and Zj ≠ Z, there is only one group of output color components. When Xi ≠ Z and Zj = Z, there is only one group of image attributes.
[0138]
As described in the first and second examples, at least one of the output component and the image attribute (selected data component) can be divided into a plurality of groups as long as Expression (5) is satisfied. In addition, since it is divided into a plurality of groups, in this case, it is assumed that at least one of the number X of output components and the number Z of selected data components is 3 or more. This is because the number of output components X or the number of selected data components Z must be plural (2 or more) in order to combine them. If X = 2 or Z = 2, only one group can be created by combining them. .
[0139]
The third example is different from the second example in that the number of input components is limited to two or more. This is because this is a combination with the first example. An example in which the number of input components is “3” or “4” is the same as in the first example. Since the specific example of the color space is the same as that of the first example, the description is omitted.
[0140]
FIG. 14 is a diagram illustrating a basic configuration example of the data conversion device 1 according to the third embodiment when an LUT having a data structure to which the synthesis method of the third example illustrated in FIG. 13 is applied is used. The differences from the configuration of the first embodiment shown in FIG. 3 or the second embodiment shown in FIG. 12 are the memory data structure in the data storage unit 2 and the function of the data separation unit 16. Hereinafter, differences will be described.
[0141]
As described with reference to FIG. 13, the data storage unit 2 of the third embodiment groups a plurality of output color components and a plurality of selected data components corresponding to a plurality of image attributes according to a predetermined standard. A plurality of basic data having the same address are combined into one piece of combined data and stored to form a look-up table LUT. For example, when all output color components and all selected data components are grouped together, as in the first example, the reference memory 2T1 for the number T of grid point data required for the interpolation calculation is used. , 2T2, ..., 2Tt. Each of the reference memories 2T1, 2T2,..., 2Tt stores grid point data for the image attributes Z1 to Zz and the output colors S1 to Sx in a combined manner.
[0142]
Each of the memories 2T1, 2T2,..., 2Tt is independently addressable so as to be able to cope with the case where the time allocated as the processing time of the color space conversion for one pixel is very short. It is configured to be readable, that is, independently readable.
[0143]
The arithmetic processing unit 4 includes, as signal processing systems, S1 signal generation units 4S1, S2 signal generation units 4S2,..., Sx signal generation units 4Sx so as to correspond to the color components S1, S2,. including. The arithmetic processing unit 4 separates the composite data stored in the data storage unit 2 into data for each image attribute (selected data component) such as a text, a computer graphics image, or a photographic image, and outputs the data. A data separation unit 16 that separates data for each component is provided. Here, in the data separation unit 16 of the third embodiment, an individual circuit is provided inside corresponding to the output colors S1, S2,..., Sx.
[0144]
The arithmetic processing unit 4 first obtains a plurality of output color components and a plurality of image attributes from grid addresses of each output color component from an arbitrary address of the memories 2T1, 2T2,. Is read out of the K-bit or Ki-bit synthesized data. The data separation unit 16 refers to a predetermined data synthesis method and converts the read synthesized data into grid point data for each image attribute such as a text, a computer graphics image, or a photographic image. The grid point data is selected according to the image attribute to be separated and processed, and is also separated into grid point data for each output color component. Since the composite data is prepared in a format summarized for each output color and image attribute, the data separation / selection process is simple. Note that the selection process of the grid point data according to the image attribute to be processed (image attribute selection process) and the separation process into the grid point data for each output color component (output color separation process) do not matter in any order. .
[0145]
The signal generators 4S1 to 4Sx provided for the respective color components S1, S2,..., Sx perform complementary operations using the grid point data separated for each output color component, thereby obtaining the color components S1, S2,. .... Generate output data for each Sx.
[0146]
Note that the processing procedure regarding color data conversion in the data conversion device 1 of the third embodiment is the same as the data separation processing in the data separation unit 16 except that separation / selection processing according to image attributes is added. This is the same as the processing procedure of the first embodiment. The flowchart showing the processing procedure is not shown.
[0147]
FIG. 15 is a diagram illustrating a specific example of a data structure when performing color conversion from RGB input to YMCK output using a multidimensional LUT in the data conversion device 1 of the third embodiment. The case of two image attributes and four output colors YMCK is shown. For comparison with the conventional example shown in FIG. 18, the number of data stored in the original (conventional) LUT is defined as 17 data × 17 data × 17 data (4913 grid point addresses are required).
[0148]
Here, as described above, the arithmetic processing unit 4 performs the tetra-hydra interpolation operation. In this case, the memory (data storage unit 2) includes four memories (RAM0 to RAM4) for four vertices so that data of four vertices necessary for the tetra-hydra interpolation operation can be simultaneously extracted within a certain unit time. RAM 3).
[0149]
At the address position of each vertex, two image attributes and grid point data for four colors of YMCK are collected and held as one composite data. Here, the RGB input data is 8 bits, and two image attributes are arranged in a row of 8 bits × 2 × 4 colors = 64 bits because two image attributes and four colors of YMCK are combined. That is, in order to perform different color conversion for each component (image object) in the output document, the data bit width is increased while maintaining the number of lattice points (vertexes) T of the LUT necessary for the interpolation operation. To maintain full capacity. According to such a configuration, the occupied area of the memory configured in the application-specific semiconductor is significantly smaller than the case where the memory is individually used for each image attribute and each output color, and the cost is reduced. The economic effect is enormous.
[0150]
FIG. 16 is a block diagram illustrating a specific configuration of the data conversion device 1 according to the third embodiment. The configuration of the data conversion device 1 shown in FIG. 16 is a specific circuit configuration to which the memory configuration shown in FIG. 5 in the case of three-dimensional input of R, G, and B is applied. Here, it is also assumed that the components of the input digital color separation signal are R, G, and B, each of which is 8-bit data.
[0151]
In the data conversion device 1 according to the third embodiment, the functions of the address generation unit 10, the tetrahedron determination unit 14, the interpolation coefficient generation unit 20, and the output color data determination unit 22 correspond to the data of the first embodiment shown in FIG. This is the same as the conversion device 1. On the other hand, the configuration of the data separation unit 16 is different from the data conversion device 1 of the first embodiment.
[0152]
Specifically, the data separation unit 16 of the third embodiment separates the combined data summarizing all the image attributes and all the output colors into each output color component, and first stores the four vertex memories 2T1 to 2T1. The first data selector 17T1, the second data selector 17T2, the third data selector 17T3, the fourth data selector 17T4, the first output color data separator 16T1, and the second output color correspond to 2T4. It includes a data separation unit 16T2, a third output color data separation unit 16T3, and a fourth output color data separation unit 16T4.
[0153]
Each of the data selection units 17T1 to 17T4 of the data separation unit 16 separates the composite data output from the memories 2T1 to 2T4 into grid point data for each image attribute such as a text, a computer graphics image, or a photographic image. Then, grid point data corresponding to the image attribute to be processed is selected and input to the corresponding output color data separation units 16T1 to 16T4. Each of the output color data separation units 16T1 to 16T4 separates the output data into unit data of the output colors Y, M, C, and K, that is, 8-bit units of interpolation target data D1, D2, D3, and D4. The data separation unit 16 inputs the interpolation target data D1, D2, D3, and D4 for each separated output color to the corresponding interpolation processing units 22Y to 22K.
[0154]
As described with reference to FIG. 14, since the order of the image attribute selection process and the output color separation process does not matter, the arrangement order of each data selection unit 17T1 to 17T4 and each output color data separation unit 16T1 to 16T4 is changed. It may be reversed.
[0155]
In the data conversion device 1 of the first embodiment, there is one neighboring data in the LUT corresponding to the input value at each vertex, and the data output by the interpolation operation is one-to-one with respect to one input. It was decided. However, it is desired that a hard copy output device such as a printer performs a different color conversion for each component (eg, text, computer graphics image, photographic image, etc.) in the output document. That is, a plurality of data is required as the data of each vertex. Which of the data of a certain vertex is to be applied to perform the subsequent interpolation operation is realized by, for example, adding additional data (tag; tag) to the input image data.
[0156]
Each of the memories 2T1 to 2T4 in the data storage unit 2 addressed by the address generation unit 10 stores two data of the Y component (Y0, Y1), two data of the M component (M0, M1), and two data of the C component. A total of eight pieces of data (C0, C1) and two pieces of K component data (K0, K1) are output as one piece of combined data. Here, the bit width is 64 bits because the original LUT grid point data is 8 bits.
[0157]
The input RGB data has, for each pixel, tag data indicating a component in the document described above. Here, it is assumed that the data is 1-bit data. Tag data is input to each of the data selection units 17T1 to 17T4. When the tag data is “0”, each of the data selecting units 17T1 to 17T4 selects the upper 32 bits of the grid point data Y0, M0, C0, and K0 of the first image attribute from the 64-bit composite data. The data of the part is separated and selected and passed to the corresponding output color data separation units 16T1 to 16T4.
[0158]
On the other hand, when the tag data is “1”, each of the data selection units 17T1 to 17T4, from the 64-bit composite data, lower-order data of the grid point data Y1, M1, C1, and K1 of the second image attribute. The data of the 32-bit portion is separated and selected, and passed to the corresponding output color data separation units 16T1 to 16T4. The output color data separation units 16T1 to 16T4 separate the interpolation target data D1, D2, D3, and D4 of the output colors Y, M, C, and K, and the interpolation target data D1, D2, and D3 for each of the separated output colors. , D4 are input to the corresponding interpolation processing units 22Y to 22K.
[0159]
The interpolation coefficients W1, W2, W3, and W4 are commonly input from the interpolation coefficient generation unit 20 to the interpolation processing units 22Y to 22K. Since the interpolation coefficients W1, W2, W3, and W4 are determined for each tetrahedron including the input point, they are not affected by the state of the tag data. As in the first embodiment, each of the interpolation processing units 22Y to 22K uses the interpolation coefficients W1, W2, W3, and W4 and the corresponding interpolation target data D1, D2, D3, and D4 to perform the tetra-hydra method. , The corresponding output color data is generated.
[0160]
As described above, the present invention has been described using the embodiment. However, the technical scope of the present invention is not limited to the scope described in the embodiment. Various changes or improvements can be made to the above-described embodiment without departing from the spirit of the invention, and embodiments with such changes or improvements are also included in the technical scope of the present invention.
[0161]
Further, the above embodiments do not limit the invention according to the claims (claims), and all combinations of the features described in the embodiments are not necessarily essential to the means for solving the invention. Absent. The embodiments described above include inventions at various stages, and various inventions can be extracted by appropriately combining a plurality of disclosed constituent features. Even if some components are deleted from all the components shown in the embodiment, as long as the effect is obtained, a configuration from which some components are deleted can be extracted as an invention.
[0162]
For example, in each of the above-described embodiments, in order to simplify the data separation processing, the composite data is prepared in a format summarized for each output color or image attribute. Or may be combined separately regardless of image attributes. At the time of the data conversion processing, the original grid point data may be extracted with reference to the synthesis method.
[0163]
Further, in the above-described embodiment, synthesized data obtained by synthesizing a plurality of pieces of grid point data is used from the viewpoint of image attributes and output colors. However, the present invention is not limited to this. Synthesized data obtained by synthesizing point data may be used.
[0164]
The applicant of the present application has proposed in Japanese Patent Application No. 2002-273976 a technique for realizing color data conversion with a smaller number of memory resources than the number of grid points (the number of data for an interpolation operation) required for an interpolation operation. I have. The method of the present embodiment described above, in which the synthesized data obtained by synthesizing the respective grid point data of the original look-up table according to a predetermined standard and storing the data as grid point data of a new look-up table, is described in Japanese Patent Application Publication No. The technique can also be applied to grid point data in the technique described in US Pat.
[0165]
The technique described in Japanese Patent Application No. 2002-273976 uses, when performing an interpolation operation in a polyhedron, a memory space that can be independently controlled in a number smaller than the number of data for the interpolation operation site, and uses each of the memory spaces. , A plurality of basic addresses are assigned to one address according to a predetermined procedure, with basic data corresponding to grid point data associated with each grid point in a k-dimensional grid space defined by k input color signal elements. The basic data for the number of data for the interpolation operation site corresponding to the pixel data to be processed is extracted from among a plurality of basic data assigned to the same address, This is to perform an interpolation operation within the polyhedron.
[0166]
If the above-described embodiment is applied to the technology described in Japanese Patent Application No. 2002-273976, the memory size can be reduced, and the number of data for the interpolation operation site is less than four (four in the case of four-point interpolation). Color data conversion can be realized with a small amount of memory resources, and power consumption and radiation noise can be reduced.
[0167]
【The invention's effect】
As is apparent from the above description, in the present invention, the combined data obtained by combining a plurality of pieces of each grid point data in the original look-up table based on a predetermined criterion are used as the grid points in the new look-up table. Store as data. The bit width of the synthesized data is increased by an amount obtained by synthesizing a plurality of data.
[0168]
When the bit width of data is increased, the gate scale (memory size) of a memory for storing data is reduced even with the same memory capacity. Therefore, by storing the synthesized data having a wide bit width in the memory (data storage unit), a memory having a smaller chip size can be used, and the memory cost can be reduced.
[Brief description of the drawings]
FIG. 1 is a diagram illustrating the relationship between the capacity of a memory, which is an example of a storage unit included in an application-specific semiconductor, and an area occupied by the memory (hereinafter referred to as a gate scale).
FIG. 2 is a diagram illustrating a first example of a data structure (memory data structure) stored in an LUT used when converting color data.
FIG. 3 is a diagram illustrating a basic configuration example of a data conversion device according to the first embodiment to which the synthesis method of the first example is applied.
FIG. 4 is a flowchart illustrating an outline of a processing procedure regarding color data conversion in the data conversion device according to the first embodiment.
FIG. 5 is a diagram illustrating a specific example of a data structure in the data conversion device according to the first embodiment.
FIG. 6 is a block diagram illustrating a specific configuration of the data conversion device according to the first embodiment.
FIG. 7 is a diagram illustrating a unit cube determination process (steps S100 to S102 in FIG. 4) in an arithmetic processing unit.
8 is a diagram illustrating a tetrahedron determination process (steps S104 to S106 in FIG. 4) in the arithmetic processing unit.
FIG. 9 is a diagram illustrating a relationship between a base point P21 and coordinate positions of four vertices based on a determination result by a tetrahedron determination unit.
FIG. 10 is a diagram illustrating an interpolation coefficient calculation process (S102c) and an interpolation calculation process (S102d) of an output value calculation process (step S108 in FIG. 4) in the calculation processing unit.
FIG. 11 is a diagram illustrating a second example of the memory data structure.
FIG. 12 is a diagram illustrating a basic configuration example of a data conversion apparatus according to a second embodiment to which the combining method of the second example is applied.
FIG. 13 is a diagram illustrating a third example of the memory data structure.
FIG. 14 is a diagram illustrating a basic configuration example of a data conversion device according to a third embodiment to which the combining method of the third example is applied.
FIG. 15 is a diagram illustrating a specific example of a data structure in the data conversion device according to the third embodiment.
FIG. 16 is a block diagram illustrating a specific configuration of a data conversion device according to a third embodiment.
FIG. 17 is a diagram illustrating a configuration example of a conventional data conversion device that performs color conversion from RGB input to YMCK output using tetra-hydra interpolation.
18 is a diagram illustrating an example of a data configuration of a reference memory of each vertex data used in a data storage unit in the data conversion device illustrated in FIG. 17;
[Explanation of symbols]
DESCRIPTION OF SYMBOLS 1 ... Data conversion device, 2 ... Data storage part, 4 ... Operation processing part, 6 ... Address bus, 8 ... Data bus, 10 ... Address generation part, 14 ... Tetrahedron determination part, 16 ... Data separation part, 20 ... Interpolation Coefficient generation unit 22, 22 output color data determination unit, 22Y, 22M, 22C, 22K interpolation processing unit

Claims (21)

For each pixel data of the input color data, determine the pixel data for the output color corresponding to the input color data based on the grid point data of the lookup table used when converting the input color data to the output color data. Data conversion method,
At each grid point of the lookup table, composite data obtained by combining a predetermined number of output colors for each grid point data in the original lookup table for the plurality of output colors based on a predetermined reference is stored.
Reading the composite data corresponding to the pixel data of the input color data,
The read composite data is separated into grid point data for the output color,
A data conversion method, wherein pixel data for the output color is determined using the separated grid point data for the output color for each of the output colors. For each pixel data of the input color data, determine the pixel data for the output color corresponding to the input color data based on the grid point data of the lookup table used when converting the input color data to the output color data. Data conversion method,
At each grid point of the lookup table, composite data obtained by combining a predetermined number of image attributes for each grid point data in the original lookup table for a plurality of image attributes based on a predetermined criterion is stored.
Reading the composite data corresponding to the pixel data of the input color data,
The read composite data is separated into grid point data for the image attributes,
A data conversion method, wherein pixel data for the output color is determined using the separated grid point data for the output color for each of the output colors. For each pixel data of the input color data, determine the pixel data for the output color corresponding to the input color data based on the grid point data of the lookup table used when converting the input color data to the output color data. Data conversion method,
At each grid point of the lookup table, each grid point data in the original lookup table for a plurality of image attributes and a plurality of output colors is stored for a predetermined number of image attributes and a predetermined number of output colors based on a predetermined reference. Store the synthesized data obtained by synthesizing
Reading the composite data corresponding to the pixel data of the input color data,
The read composite data is separated into each grid point data for the image attribute and the output color,
A data conversion method, wherein pixel data for the output color is determined using the separated grid point data for the output color for each of the output colors. For each pixel data of the input color data, it is determined to which of the unit lattices constituting the set of the lattice point data the pixel data belongs, and the corresponding unit lattice is further divided into a plurality of polyhedrons. Then, it is determined which of the polyhedrons the pixel data belongs to, and an output pixel data corresponding to the input pixel data is determined by an interpolation operation in the corresponding polyhedron. The data conversion method according to any one of claims 1 to 3. For each pixel data of the input color data, determine the pixel data for the output color corresponding to the input color data based on the grid point data of the lookup table used when converting the input color data to the output color data. Data conversion device,
A data storage in which composite data obtained by synthesizing each grid point data in the original lookup table for the plurality of output colors with a predetermined number of output colors based on a predetermined standard is stored at each grid point. Department and
A combined data reading unit that reads the combined data corresponding to the pixel data of the input color data from the data storage unit;
A data separation unit that separates the combined data read by the combined data reading unit into respective grid point data for the output color;
An output color data determining unit that determines pixel data for the output color using grid point data for the output color separated by the data separating unit for each of the output colors. Data converter. The predetermined criterion is N = X, where X is the number of components of the plurality of output colors, Y is the bit width of each grid point data in the original lookup table, and N is the bit width of the composite data. 6. The data conversion device according to claim 5, wherein Y is set. The predetermined criterion is such that when the components of the plurality of output colors are divided into a plurality of groups, the number of components of the output color belonging to the i-th group is xi, and the number of bits of the composite data in the i-th group is 6. The data conversion device according to claim 5, wherein when the width is Ni, Ni = xi.Y. For each pixel data of the input color data, determine the pixel data for the output color corresponding to the input color data based on the grid point data of the lookup table used when converting the input color data to the output color data. Data conversion device,
A data storage unit in which combined data obtained by combining each grid point data in the original look-up table for a plurality of image attributes into a predetermined number of image attributes based on a predetermined reference is stored at each grid point; ,
A combined data reading unit that reads the combined data corresponding to the pixel data of the input color data from the data storage unit;
A data separating unit that separates the combined data read by the combined data reading unit into respective grid point data for the image attributes;
An output color data determining unit that determines pixel data for the output color using grid point data for the output color separated by the data separating unit for each of the output colors. Data converter. The predetermined criterion is that when the number of components of the image attribute is Z, the bit width of each grid point data in the original lookup table is Y, and the bit width of the composite data is L, L = Z · Y 9. The data conversion device according to claim 8, wherein: The predetermined criterion is that when the components of the plurality of image attributes are divided into a plurality of groups, the number of components of the image attributes belonging to the j-th group is zj, and the number of bits of the combined data in the j-th group is 9. The data conversion device according to claim 8, wherein when the width is Lj, Lj = zj.Y. For each pixel data of the input color data, determine the pixel data for the output color corresponding to the input color data based on the grid point data of the lookup table used when converting the input color data to the output color data. Data conversion device,
Synthesized data obtained by synthesizing each grid point data in the original look-up table for a plurality of image attributes and a plurality of output colors into a predetermined number of image attributes and a predetermined number of output colors based on a predetermined reference is obtained. A data storage unit stored at each grid point,
A combined data reading unit that reads the combined data corresponding to the pixel data of the input color data from the data storage unit;
A data separating unit that separates the combined data read by the combined data reading unit into respective grid point data for the image attribute and the output color;
An output color data determining unit that determines pixel data for the output color using grid point data for the output color separated by the data separating unit for each of the output colors. Data converter. The predetermined criterion is that the number of components of the plurality of output colors is X, the number of components of the image attribute is Z, the bit width of each grid point data in the original lookup table is Y, and the bit of the composite data is 12. The data conversion device according to claim 11, wherein when the width is K, K = XZZY. The predetermined criterion is such that when the components of the plurality of output colors are divided into a plurality of groups, the number of components of the output color belonging to the i-th group is xi, the number of components of the image attribute is Z, 12. The data conversion device according to claim 11, wherein when a bit width of the combined data in the i-th group is Ki, Ki = xi.Z.Y. The predetermined criterion is that the number of components of the plurality of output colors is X, and the number of components of the image attribute belonging to the j-th group when the components of the plurality of image attributes are divided into a plurality of groups is zj. 12. The data conversion apparatus according to claim 11, wherein when a bit width of the combined data in the j-th group is Kj, Kj = X · zj · Y. The predetermined criterion is such that when the components of the plurality of output colors are divided into a plurality of groups, the number of components of the output color belonging to the i-th group is xi, and the components of the plurality of image attributes are a plurality of components. When the number of components of the image attribute belonging to the j-th group when divided into groups is zj, and the bit width of the composite data in the j-th group is Kij, Kij = xi · zj · Y The data conversion device according to claim 11, wherein For each pixel data of the input color data, determine the pixel data for the output color corresponding to the input color data based on the grid point data of the lookup table used when converting the input color data to the output color data. Data conversion device,
A data storage unit in which combined data obtained by combining a plurality of pieces of each grid point data in the original lookup table based on a predetermined reference is stored at each grid point,
A combined data reading unit that reads the combined data corresponding to the pixel data of the input color data from the data storage unit;
A data separation unit that separates the combined data read by the combined data reading unit into respective grid point data of the original lookup table;
An output color data determining unit that determines pixel data for the output color using grid point data for the output color separated by the data separating unit for each of the output colors. Data converter. The data conversion device according to any one of claims 5 to 7, 11 to 15, wherein the number of components of the input color is two. 11. The data conversion device according to claim 8, wherein the number of components of the input color is one. The number of components of the input color is three, and the color space defined by the components of the input color is an RGB color space, an sRGB color space, a Lab color space, a YCbCr color space, a YMC color space, and an XYZ color space. The data conversion device according to any one of claims 5 to 16, wherein the data conversion device is any one of the following. 17. The method according to claim 5, wherein the number of components of the input color is four, and the color spaces defined by the components of the input color are a YMCK color space, a K + Lab color space, and an RGB + L color space. The data conversion device according to any one of the preceding claims. For each pixel data of the input color data, a unit lattice body determining unit that determines which of the unit lattices constituting the set of the lattice point data the pixel data belongs to,
A polyhedron determination unit that further divides the corresponding unit lattice determined by the unit lattice determination unit into a plurality of polyhedrons and determines which of the polyhedrons the pixel data belongs to,
The output color data determination unit may include an interpolation processing unit that determines output pixel data corresponding to the input pixel data by an interpolation operation in the corresponding polyhedron determined by the polyhedron determination unit. The data conversion device according to any one of claims 5 to 20, wherein:

Images