JP2004350072A - Image coder, image coding method, image coding program and recording medium with the program recorded thereon - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To realize quality enhancement of a decoded image by performing coding according to characteristics of a residual signal by simple structure when hierarchical image coding is performed. <P>SOLUTION: When extended layer coding data is generated by performing coding for the prescribed number of planes from the most significant bit plane to less significant bit planes, absolute values of orthogonal conversion factors having values in the most significant bit plane are made small by one most significant bit and the coding for the prescribed number of planes is performed by considering that there are few orthogonal conversion factors having the values in the most significant bit plane. In addition, in the case of judging that there are few orthogonal conversion factors having the values in the most significant bit plane and judging that there are substantially few orthogonal conversion factors having the values in the most significant bit plane, the absolute values of the orthogonal conversion factors having the values in the most significant bit plane are made small by one most significant bit and the coding for the prescribed number of planes is performed. <P>COPYRIGHT: (C)2005,JPO&NCIPI

Description

で定義される関数であって、ｘの符号を表している。
【００４３】
例えば、ｍａｘ＝６である場合には、係数処理部４０１は、
６４＝２^６ ≦｜ｃｏｅｆ｜≦２^６＋１ −１＝１２７
という値域にあるＤＣＴ係数ｃｏｅｆを、
ｃｏｅｆ＿ｃｕｔ＝２^６ −１＝６３
に置き換える処理を実行することになる。
【００４４】
すなわち、係数処理部４０１は、符号化対象のＤＣＴ係数の絶対値を最上位１ビット分だけ小さくするように処理するのである。以下、この係数処理部４０１の実行する処理を“頭打ち処理（ＤＣＴ係数の頭打ち処理）”と呼ぶことにする。
【００４５】
次に、図２の処理フローに従って、図１のように構成される第２の符号化器１０６の実行する符号化処理について詳細に説明する。
【００４６】
第２の符号化器１０６は、図２の処理フローに示すように、先ず最初に、ステップ１０で、符号化対象となる差分画像信号を入力すると、続いて、ステップ１１で、その差分画像信号に対して、マクロブロック毎に、離散コサイン変換を施すことでＤＣＴ係数を算出する。
【００４７】
続いて、ステップ１２で、マクロブロック毎に、算出したＤＣＴ係数に対してビットシフト処理を施す。
【００４８】
続いて、ステップ１３で、フレーム全体のＤＣＴ係数のうち絶対値が最大のものを検索することで、ビットプレーン数（上述のｍａｘ）を決定する。例えば、６５というＤＣＴ係数の絶対値が最大値である場合には、ビットプレーン数としてｍａｘ＝６であることを決定するのである。
【００４９】
続いて、ステップ１４で、フレームの全マクロブロックについて処理を終了したのか否かを判断して、未処理のマクロブロックが残されていることを判断するときには、ステップ１５に進んで、未処理のマクロブロックを１つ選択する。
【００５０】
続いて、ステップ１６で、選択したマクロブロックについて、最上位ビットプレーンに値を持つＤＣＴ係数に対して上述の頭打ち処理を施してから、ステップ１４に戻る。例えば、６５という絶対値を持つＤＣＴ係数については、その絶対値が６３となるように置換する頭打ち処理を施してから、ステップ１４に戻るのである。
【００５１】
このようにして、ステップ１４ないしステップ１６の処理を繰り返していくときに、ステップ１４で、フレームの全マクロブロックについて処理を終了したことを判断すると、ステップ１７に進んで、最上位のビットプレーンから下位のビットプレーンに向けて、ＦＧＳで規定されるビットプレーン数分可変長符号化を行って、処理を終了する。具体的には、上位４プレーンに対して可変長符号化を行って、処理を終了するのである。
【００５２】
次に、このように構成される本発明の符号化処理の有効性について、具体的な計算例を用いて検証する。
【００５３】
図３（ａ）に示す８×８画素の画像差分値に対して離散コサイン変換を施すと、図３（ｂ）に示すＤＣＴ係数が得られることになる。
【００５４】
従来技術の符号化処理に従う場合には、図３（ｂ）に示すＤＣＴ係数の上位４プレーンに対して可変長符号化を行い、それを復号することで図３（ｄ）に示す復号差分値を得ることになる。
【００５５】
これに対して、本発明の符号化処理では、図３（ｂ）に示すＤＣＴ係数に対して上述の頭打ち処理を施すことで図３（ｃ）に示すＤＣＴ係数に変換し、その変換したＤＣＴ係数の上位４プレーンに対して可変長符号化を行い、それを復号することで図３（ｅ）に示す復号差分値を得ることになる。すなわち、上述の頭打ち処理に従って、６５という絶対値を持つＤＣＴ係数については６３という絶対値を持つものに変換しつつ、可変長符号化を行うことになる。
【００５６】
図３（ａ）に示す画像差分値と図３（ｄ）に示す復号差分値とから、従来技術の符号化処理に従う場合には、最大誤差が１２で、誤差の総和は２０８となる。これに対して、図３（ａ）に示す画像差分値と図３（ｅ）に示す復号差分値とから分かるように、本発明の符号化処理によれば、最大誤差が７で、誤差の総和は１３３となる。
【００５７】
このように、本発明の符号化処理によれば、復号画像の品質の向上を実現できるようになるのである。
【００５８】
係数処理部４０１は、図２の処理フローに従う場合、最上位ビットプレーンに対して頭打ち処理を１回施すことで説明したが、最上位１ビット分だけ小さくしたＤＣＴ係数を新たな処理対象として、その新たな処理対象の最上位のビットプレーンに値を持つＤＣＴ係数の個数に応じて、さらに頭打ち処理を継続していくように処理することも可能である。
【００５９】
図４に示す処理フローに従って、この構成を実現する第２の符号化器１０６の実行する符号化処理について詳細に説明する。
【００６０】
第２の符号化器１０６は、図４の処理フローに従う場合には、先ず最初に、ステップ２０で、符号化対象となる差分画像信号を入力すると、続いて、ステップ２１で、その差分画像信号に対して、マクロブロック毎に、離散コサイン変換を施すことでＤＣＴ係数を算出する。
【００６１】
続いて、ステップ２２で、マクロブロック毎に、算出したＤＣＴ係数に対してビットシフト処理を施し、続くステップ２３で、フレーム全体のＤＣＴ係数のうち絶対値が最大のものを検索することで、ビットプレーン数を決定する。
【００６２】
続いて、ステップ２４で、フレームの全マクロブロックについて処理を終了したのか否かを判断して、未処理のマクロブロックが残されていることを判断するときには、ステップ２５に進んで、未処理のマクロブロックを１つ選択し、続くステップ２６で、選択したマクロブロックについて、最上位ビットプレーンに値を持つＤＣＴ係数に対して頭打ち処理を施してから、ステップ２４に戻る。
【００６３】
このようにして、ステップ２４ないしステップ２６の処理を繰り返していくときに、ステップ２４で、フレームの全マクロブロックについて処理を終了したことを判断すると、ステップ２７に進んで、頭打ち処理を施したＤＣＴ係数を、フレームの持つ新たなＤＣＴ係数として設定し、続くステップ２８で、フレーム全体のＤＣＴ係数のうち絶対値が最大のものを検索することで、ビットプレーン数を決定する。
【００６４】
続いて、ステップ２９で、最上位のビットプレーンに値を持つＤＣＴ係数のフレーム全体に対する割合が規定の閾値よりも大きいのか否かを判断して、小さいことを判断するときには、新たなＤＣＴ係数に対して頭打ち処理を継続すべくステップ２４に戻る。
【００６５】
一方、ステップ２９で、最上位のビットプレーンに値を持つＤＣＴ係数のフレーム全体に対する割合が規定の閾値よりも大きいことを判断するときには、頭打ち処理を継続するのは適切でないと判断して、ステップ３０に進んで、最上位のビットプレーンから下位のビットプレーンに向けて、ＦＧＳで規定されるビットプレーン数分可変長符号化を行って、処理を終了する。
【００６６】
このようにして、係数処理部４０１は、図４の処理フローに従う場合には、最上位１ビット分だけ小さくしたＤＣＴ係数を新たな処理対象として、その新たな処理対象の最上位のビットプレーンに値を持つＤＣＴ係数の個数に応じて、さらに頭打ち処理を継続していくように処理するのである。
【００６７】
図５に、本発明の他の実施形態例を図示する。
【００６８】
ここで、図中、図１に示したものと同じのものについては同一の記号で示してある。
【００６９】
この実施形態例に従う場合、本発明により構成される拡張レイヤを符号化する第２の符号化器１０６は、ＦＧＳ符号化における階層的符号化方式を実装するにあたって、新たに係数カウント・判定部４０２を備えることを特徴とする。
【００７０】
この係数カウント・判定部４０２は、最上位ビットプレーンに含まれるＤＣＴ係数のフレーム全体に対する割合ｐを求め、その割合ｐに従って頭打ち処理を実行するのか否かを判定する処理を実行する。
【００７１】
次に、図６の処理フローに従って、図５のように構成される第２の符号化器１０６の実行する符号化処理について詳細に説明する。
【００７２】
第２の符号化器１０６は、図５のように構成される場合には、図６の処理フローに示すように、先ず最初に、ステップ４０で、符号化対象となる差分画像信号を入力すると、続いて、ステップ４１で、その差分画像信号に対して、マクロブロック毎に、離散コサイン変換を施すことでＤＣＴ係数を算出する。
【００７３】
続いて、ステップ４２で、マクロブロック毎に、算出したＤＣＴ係数に対してビットシフト処理を施し、続くステップ４３で、フレーム全体のＤＣＴ係数のうち絶対値が最大のものを検索することで、ビットプレーン数を決定する。
【００７４】
続いて、ステップ４４で、最上位のビットプレーンに値を持つＤＣＴ係数のフレーム全体に対する割合が規定の閾値よりも大きいのか否かを判断して、大きいことを判断するときには、頭打ち処理を行うのは適切でないと判断して、ステップ４５に進んで、最上位のビットプレーンから下位のビットプレーンに向けて、ＦＧＳで規定されるビットプレーン数分可変長符号化を行って、処理を終了する。
【００７５】
すなわち、最上位のビットプレーンに値を持つＤＣＴ係数のフレーム全体に対する割合が規定の閾値よりも大きい場合には、頭打ち処理を行わずに、最上位のビットプレーンから下位のビットプレーンに向けて、ＦＧＳで規定されるビットプレーン数分可変長符号化を行うのである。
【００７６】
一方、ステップ４４で、最上位のビットプレーンに値を持つＤＣＴ係数のフレーム全体に対する割合が規定の閾値よりも大きくないことを判断するときには、ステップ４６に進んで、フレームの全マクロブロックについて処理を終了したのか否かを判断して、未処理のマクロブロックが残されていることを判断するときには、ステップ４７に進んで、未処理のマクロブロックを１つ選択する。
【００７７】
続いて、ステップ４８で、選択したマクロブロックについて、最上位ビットプレーンに値を持つＤＣＴ係数に対して頭打ち処理を施す。
【００７８】
このようにして、ステップ４６ないしステップ４８の処理を繰り返していくときに、ステップ４６で、フレームの全マクロブロックについて処理を終了したことを判断すると、ステップ４５に進んで、最上位のビットプレーンから下位のビットプレーンに向けて、ＦＧＳで規定されるビットプレーン数分可変長符号化を行って、処理を終了する。
【００７９】
すなわち、最上位のビットプレーンに値を持つＤＣＴ係数のフレーム全体に対する割合が規定の閾値よりも小さい場合には、ＤＣＴ係数に対して頭打ち処理を施して、その頭打ち処理を施したＤＣＴ係数を符号化対象として、最上位のビットプレーンから下位のビットプレーンに向けて、ＦＧＳで規定されるビットプレーン数分可変長符号化を行うのである。
【００８０】
このようにして、第２の符号化器１０６は、図５のように構成される場合には、最上位ビットプレーンに含まれるＤＣＴ係数のフレーム全体に対する割合ｐを求め、その割合ｐとあらかじめ設定した閾値Ｔｈ１とを比較して、ｐ≦Ｔｈ１ならば、頭打ち処理を施すように処理するのである。
【００８１】
この構成に従って、頭打ち処理の対象となるＤＣＴ係数の割合が多い場合に発生することになる、復号された残差信号の精度が劣化するという問題を回避することができるようになる。
【００８２】
係数カウント・判定部４０２は、図６の処理フローに従う場合、最上位ビットプレーンに値を持つＤＣＴ係数の分布については考慮していないが、この分布を考慮して、頭打ち処理を実行するのか否かを判定するようにするという構成を採ることも可能である。
【００８３】
図７に示す処理フローに従って、この構成を実現する第２の符号化器１０６の実行する符号化処理について詳細に説明する。
【００８４】
第２の符号化器１０６は、図７の処理フローに従う場合には、先ず最初に、ステップ５０で、符号化対象となる差分画像信号を入力すると、続いて、ステップ５１で、その差分画像信号に対して、マクロブロック毎に、離散コサイン変換を施すことでＤＣＴ係数を算出する。
【００８５】
続いて、ステップ５２で、マクロブロック毎に、算出したＤＣＴ係数に対してビットシフト処理を施し、続くステップ５３で、フレーム全体のＤＣＴ係数のうち絶対値が最大のものを検索することで、ビットプレーン数を決定する。
【００８６】
続いて、ステップ５４で、最上位のビットプレーンに値を持つＤＣＴ係数のうちの値が大きいものの個数を判断する。
【００８７】
例えば、ある閾値Ｔｈ２に対して、
２^ｍａｘ ≦｜ｃｏｅｆ｜≦２^ｍａｘ ＋Ｔｈ２
という条件を満たすｃｏｅｆの集合を｛ＣＯＥＦ２｝と表し、
２^ｍａｘ ＋Ｔｈ２＜｜ｃｏｅｆ｜＜２^{ｍａｘ＋１}
という条件を満たすｃｏｅｆの集合を｛ＣＯＥＦ３｝と表すときに、その｛ＣＯＥＦ３｝に属するｃｏｅｆの個数Ｎ（ＣＯＥＦ３）が、ある閾値Ｔｈ３に対して、
Ｎ（ＣＯＥＦ３）≦Ｔｈ３
であるのか否かを判断することで、最上位のビットプレーンに値を持つＤＣＴ係数のうちの値が大きいものの個数を判断するのである。
【００８８】
続いて、ステップ５５で、ステップ５４での判断処理により、最上位のビットプレーンに値を持つＤＣＴ係数のうちの値が大きいものの個数が多いという判断結果が得られたのか否かを判断して、多いという判断結果が得られたときには、ステップ５６に進んで、最上位のビットプレーンから下位のビットプレーンに向けて、ＦＧＳで規定されるビットプレーン数分可変長符号化を行って、処理を終了する。
【００８９】
すなわち、最上位のビットプレーンに値を持つＤＣＴ係数のうちの値が大きいものの個数が多い場合には、頭打ち処理を行わずに、最上位のビットプレーンから下位のビットプレーンに向けて、ＦＧＳで規定されるビットプレーン数分可変長符号化を行うのである。
【００９０】
一方、ステップ５５で、最上位のビットプレーンに値を持つＤＣＴ係数のうちの値が大きいものの個数が少ないことを判断するときには、ステップ５７に進んで、フレームの全マクロブロックについて処理を終了したのか否かを判断して、未処理のマクロブロックが残されていることを判断するときには、ステップ５８に進んで、未処理のマクロブロックを１つ選択する。
【００９１】
続いて、ステップ５９で、選択したマクロブロックについて、最上位ビットプレーンに値を持つＤＣＴ係数に対して頭打ち処理を施す。
【００９２】
このようにして、ステップ５７ないしステップ５９の処理を繰り返していくときに、ステップ５７で、フレームの全マクロブロックについて処理を終了したことを判断すると、ステップ５６に進んで、最上位のビットプレーンから下位のビットプレーンに向けて、ＦＧＳで規定されるビットプレーン数分可変長符号化を行って、処理を終了する。
【００９３】
すなわち、最上位のビットプレーンに値を持つＤＣＴ係数のうちの値が大きいものの個数が少ない場合には、ＤＣＴ係数に対して頭打ち処理を施して、その頭打ち処理を施したＤＣＴ係数を符号化対象として、最上位のビットプレーンから下位のビットプレーンに向けて、ＦＧＳで規定されるビットプレーン数分可変長符号化を行うのである。
【００９４】
このようにして、係数カウント・判定部４０２は、図７の処理フローに従う場合には、最上位のビットプレーンに値を持つＤＣＴ係数のうちの値が大きいものの個数Ｎ（ＣＯＥＦ３）と、あらかじめ設定した閾値Ｔｈ３とを比較して、Ｎ（ＣＯＥＦ３）≦Ｔｈ３ならば、頭打ち処理を施すように処理するのである。
【００９５】
この構成に従って、ＤＣＴ係数の頭打ち処理による値の変化が大きいときに発生することになる、復号された残差信号の精度が劣化するという問題を回避することができるようになる。
【００９６】
図８に、本発明の他の実施形態例を図示する。
【００９７】
ここで、図中、図５に示したものと同じのものについては同一の記号で示してある。
【００９８】
この実施形態例に従う場合、係数カウント・判定部４０２は、第１の符号化器１０１を構成する制御部２０６（図１１に示したもの）から、信号線４０３を介して符号化パラメータを受け取って、この符号化パラメータを使って、頭打ち処理を実行するのか否かを判定する処理を行うことを特徴とする。
【００９９】
例えば、原画像の複雑な領域に対応するマクロブロックでは第一の符号化器１０１における量子化パラメータが大きくなりやすいことを利用して、係数カウント・判定部４０２は、第１の符号化器１０１を構成する制御部２０６から量子化パラメータＱＰを受け取って、
Ｎ（ＣＯＥＦ４）≦Ｔｈ５
であるのか否かを判断して、この条件が成立する場合に、頭打ち処理を実行するように判断する。
【０１００】
ここで、集合｛ＣＯＥＦ４｝は、上述の集合｛ＣＯＥＦ３｝の要素ｘのうち、ＱＰｘ ≦Ｔｈ４を満たすものとして定義される。ただし、ＱＰｘ は、ＤＣＴ係数ｘの含まれるマクロブロックのベースレイヤにおける量子化パラメータであり、Ｎ｛ＣＯＥＦ４｝は、集合｛ＣＯＥＦ４｝に属するＤＣＴ係数の個数である。
【０１０１】
次に、図９の処理フローに従って、図８のように構成される第２の符号化器１０６の実行する符号化処理について詳細に説明する。
【０１０２】
第２の符号化器１０６は、図８のように構成される場合には、図９の処理フローに示すように、先ず最初に、ステップ６０で、符号化対象となる差分画像信号を入力すると、続いて、ステップ６１で、その差分画像信号に対して、マクロブロック毎に、離散コサイン変換を施すことでＤＣＴ係数を算出する。
【０１０３】
続いて、ステップ６２で、マクロブロック毎に、算出したＤＣＴ係数に対してビットシフト処理を施し、続くステップ６３で、フレーム全体のＤＣＴ係数のうち絶対値が最大のものを検索することで、ビットプレーン数を決定する。
【０１０４】
続いて、ステップ６４で、最上位のビットプレーンに値を持つＤＣＴ係数のうちの値が大きいものの中から、
Ｎ（ＣＯＥＦ４）≦Ｔｈ５
に従って、量子化パラメータが小さなものの個数を判断する。
【０１０５】
続いて、ステップ６５で、ステップ６４での判断処理により、Ｎ（ＣＯＥＦ４）≦Ｔｈ５という条件が成立したのか否かを判断して、この条件が成立しないという判断結果が得られたときには、ステップ６６に進んで、最上位のビットプレーンから下位のビットプレーンに向けて、ＦＧＳで規定されるビットプレーン数分可変長符号化を行って、処理を終了する。
【０１０６】
すなわち、Ｎ（ＣＯＥＦ４）≦Ｔｈ５という条件が成立しない場合、したがってＮ（ＣＯＥＦ４）の個数が多い場合には、頭打ち処理を行わずに、最上位のビットプレーンから下位のビットプレーンに向けて、ＦＧＳで規定されるビットプレーン数分可変長符号化を行うのである。
【０１０７】
一方、ステップ６５で、Ｎ（ＣＯＥＦ４）≦Ｔｈ５という条件が成立したことを判断するときは、ステップ６７に進んで、フレームの全マクロブロックについて処理を終了したのか否かを判断して、未処理のマクロブロックが残されていることを判断するときには、ステップ６８に進んで、未処理のマクロブロックを１つ選択する。
【０１０８】
続いて、ステップ６９で、選択したマクロブロックについて、最上位ビットプレーンに値を持つＤＣＴ係数に対して頭打ち処理を施す。
【０１０９】
このようにして、ステップ６７ないしステップ６９の処理を繰り返していくときに、ステップ６７で、フレームの全マクロブロックについて処理を終了したことを判断すると、ステップ６６に進んで、最上位のビットプレーンから下位のビットプレーンに向けて、ＦＧＳで規定されるビットプレーン数分可変長符号化を行って、処理を終了する。
【０１１０】
すなわち、Ｎ（ＣＯＥＦ４）≦Ｔｈ５という条件が成立した場合、したがってＮ（ＣＯＥＦ４）の個数が少ない場合には、ＤＣＴ係数に対して頭打ち処理を施して、その頭打ち処理を施したＤＣＴ係数を符号化対象として、最上位のビットプレーンから下位のビットプレーンに向けて、ＦＧＳで規定されるビットプレーン数分可変長符号化を行うのである。
【０１１１】
このようにして、係数カウント・判定部４０２は、図９の処理フローに従う場合には、上述の集合｛ＣＯＥＦ３｝の要素ｘのうち、ＱＰｘ ≦Ｔｈ４を満たすものとして定義される集合｛ＣＯＥＦ４｝の個数Ｎ（ＣＯＥＦ４）と、あらかじめ設定した閾値Ｔｈ５とを比較して、Ｎ（ＣＯＥＦ４）≦Ｔｈ５ならば、頭打ち処理を施すように処理するのである。
【０１１２】
この構成に従って、複雑な領域では復号画像の誤差が目立ちにくいという視覚特性を利用して、そのような領域に属するＤＣＴ係数については、上述したＮ（ＣＯＥＦ３）≦Ｔｈ３の判定から除くようにするのである。
【０１１３】
これから、ＤＣＴ係数の頭打ち処理による値の変化が大きいときに発生することになる、復号された残差信号の精度が劣化するという問題を確実に回避することができるようになる。
【０１１４】
図示実施形態例について本発明を説明したが、本発明はこれに限定されるものではない。例えば、図５及び図８に示した実施形態例では説明しなかったが、図５及び図８に示した実施形態例においても、最上位１ビット分だけ小さくしたＤＣＴ係数を新たな処理対象として、その新たな処理対象の最上位のビットプレーンに値を持つＤＣＴ係数の個数に応じて、さらに頭打ち処理を継続していくように処理するという図４の処理フローで説明した構成を用いるようにしてもよいのである。
【０１１５】
【発明の効果】
以上説明したように、本発明では、最上位のビットプレーンから下位のビットプレーンに向けて規定のプレーン数分符号化することで拡張レイヤ符号化データを生成するときに、通常の場合、最上位のビットプレーンに値を持つ直交変換係数が少ないことを考慮して、最上位のビットプレーンに値を持つ直交変換係数の絶対値を最上位１ビット分だけ小さくして、その規定のプレーン数分の符号化を行うようにする構成を採ることから、簡単な構成で残差信号の特性に応じた符号化を行い復号画像の品質の向上を実現できるようになる。
【０１１６】
そして、本発明では、最上位のビットプレーンから下位のビットプレーンに向けて規定のプレーン数分符号化することで拡張レイヤ符号化データを生成するときに、最上位のビットプレーンに値を持つ直交変換係数が少ないことを判断したり、最上位のビットプレーンに値を持つ直交変換係数が実質的に少ないことを判断する場合には、最上位のビットプレーンに値を持つ直交変換係数の絶対値を最上位１ビット分だけ小さくして、その規定のプレーン数分の符号化を行うようにする構成を採ることから、簡単な構成で残差信号の特性に応じた符号化を行い復号画像の品質の向上を実現できるようになる。
【図面の簡単な説明】
【図１】本発明の一実施形態例である。
【図２】本発明の実行する処理フローの一例である。
【図３】本発明の符号化処理の有効性について説明する図である。
【図４】本発明の実行する処理フローの一例である。
【図５】本発明の他の実施形態例である。
【図６】本発明の実行する処理フローの一例である。
【図７】本発明の実行する処理フローの一例である。
【図８】本発明の他の実施形態例である。
【図９】本発明の実行する処理フローの一例である。
【図１０】階層的符号化処理の説明図である。
【図１１】階層的符号化処理の説明図である。
【図１２】階層的符号化処理の説明図である。
【図１３】ＤＣＴ係数の説明図である。
【図１４】ＦＧＳ符号化処理の説明図である。
【符号の説明】
１０５ 差分画像信号
１０６ 第２の符号化器
１０８ 拡張レイヤ符号化データ
３０１ 離散コサイン変換処理部
３０２ 重み付け部
３０３ 最大値検索部
３０４ 可変長符号化部
４０１ 係数処理部[0001]
TECHNICAL FIELD OF THE INVENTION
The present invention provides an image encoding apparatus and a method thereof that can be used in hierarchical image encoding to perform encoding according to characteristics of a residual signal with a simple configuration and to improve the quality of a decoded image. The present invention relates to an image encoding program used for realizing the image encoding method and a recording medium on which the program is recorded.
[0002]
[Prior art]
As a method of efficiently encoding an image, there is a hierarchical image encoding method.
[0003]
This method provides a first encoder and a second encoder, and decodes encoded data generated for each small area called a macroblock by the first encoder. Obtaining a base layer decoded image of an image, and using both the encoded data generated by the first encoder and the enhancement layer encoded data generated for each macroblock by the second encoder. Thus, a decoded image with higher image quality can be obtained.
[0004]
As a method of realizing such a hierarchical encoding, a method of encoding a difference between an original image and a decoded image of encoded data obtained as a result of the first encoding by a second encoder is considered. ing. FIG. 10 shows a configuration example of this method.
[0005]
That is, in this method, as shown in FIG. 10, the base layer coded data 107 is generated by coding the original image signal in the first coder 101. Further, the first decoder 102 decodes the base layer encoded data to generate a base layer decoded image signal 103, and the subtractor 104 subtracts the base layer decoded image signal 103 from the original image signal to generate a difference image signal 105. Generate Then, the difference image signal 105 is subjected to variable-length encoding such as Huffman encoding in the second encoder 106, so that the enhancement layer encoded data 108 is generated.
[0006]
At this time, the difference image signal 105 is generally divided into a position and a size corresponding to the macroblock of the base layer and encoded.
[0007]
As a document describing the above-mentioned hierarchical coding method, there is, for example, Patent Document 1 shown below. According to the technology described in Patent Document 1, encoded data is given a hierarchy, and only a selected area is reproduced with various image qualities from a part of the encoded data, or a low-quality image is reproduced from a part of the encoded data. Or play it.
[0008]
On the other hand, in MPEG-4 coding (document: ISO / IEC 14496-2 Coding of Audio-Visual Objects, Part-2), which is an international standard for video coding, FGS is one of the hierarchical coding methods. (Fine Granularity Scalable) coding is specified (for example, see Non-Patent Document 1).
[0009]
The hierarchical coding scheme in FGS coding is as follows.
[0010]
As shown in FIG. 11, the first encoder 101 includes an MPEG-4 encoder 200 having a standard configuration of MPEG-4. In this configuration, the above-described first decoder 102 is included (2b shown in the figure). The encoding process of MPEG-4 in the MPEG-4 encoder 200 has been introduced in various documents and will not be described here.
[0011]
FIG. 12 shows a configuration example of the second encoder 106 that encodes the enhancement layer.
[0012]
In the second encoder 106, first, a discrete cosine transform processing unit 301 performs a discrete cosine transform (DCT) on the above-described difference image signal 105 to convert a pixel value difference into a frequency space. Subsequently, the weighting unit 302 performs a bit shift process on the DCT coefficients calculated by the discrete cosine transform processing unit 301 for each macroblock or for each frequency component.
[0013]
After that, the maximum value search unit 303 finds the DCT coefficient having the largest absolute value among the DCT coefficients of the entire frame (entire image). Finally, the variable-length coding unit 304 performs variable-length coding from the highest-order bit plane including the maximum value of the DCT coefficient to the lower-order bit plane.
[0014]
Here, attention is paid to the distribution range of the DCT coefficient. Since the power tends to concentrate on a specific frequency component, which is a feature of DCT, the DCT coefficient for the pixel value difference may have a large coefficient value range (dynamic range). For example, the DCT coefficient for the pixel value difference whose range shown in FIG. 13A is −56 to 46 is −65 to 49 as shown in FIG. I understand.
[0015]
In the FGS encoding method, the number of bit planes necessary for encoding the maximum absolute value of the DCT coefficient in the enhancement layer is represented by vop-max-level. The absolute value of the DCT coefficient is encoded in the zigzag scan order of the small block from the most significant bit (MSB; Most Significant Bit) of the bit plane indicated by vop-max-level. Also, the sign of the DCT coefficient is encoded together in the bit plane in which the DCT coefficient is encoded first.
[0016]
FIG. 14 shows an example of encoding in the FGS encoding method. In the example of this figure, six bit planes are assumed as vop-max-level.
[0017]
Here, if the maximum absolute value of the DCT coefficient in a certain small block does not reach the current bit plane to be coded, a flag “msb-not-reached” indicating that the DCT coefficient has not reached yet is set, Send.
[0018]
[Patent Document 1]
JP-A-2002-44671
[Non-patent document 1]
ISO / IEC 14496-2: 1999 / FDAM4, Amendment 4
[0019]
[Problems to be solved by the invention]
However, according to the conventional FGS coding method, there is a problem that the quality of a decoded image is deteriorated.
[0020]
That is, when the number of DCT coefficients to be coded in the MSB plane (most significant bit plane) indicated by vop-max-level is small, the above-mentioned msb-not-reached is coded in a large amount. The code amount of the coefficient decreases, and the quality of the decoded image deteriorates.
[0021]
From preliminary experiments by the inventor, it has been found that the ratio of the DCT coefficient included in the MSB plane is as small as about 0.3% of the entire image, and about 7% or less in a macroblock unit of 16 × 16 pixels.
[0022]
On the other hand, according to the definition of the FGS profile, the maximum number of bit planes that can be transmitted in the enhancement layer is four.
[0023]
From these facts, it is considered that the accuracy of the decoded difference value can be improved by adjusting the DCT coefficients transmitted with the same number of encoding planes.
[0024]
The present invention has been made in view of such circumstances, and is used in hierarchical image coding to improve the quality of a decoded image by performing coding according to characteristics of a residual signal with a simple configuration. The purpose of the present invention is to provide a new image coding technology that can realize the above.
[0025]
[Means for Solving the Problems]
In order to achieve the above object, an image encoding apparatus according to the present invention is used in hierarchical image encoding, wherein a difference image signal between an original image signal and a base layer decoded image signal is input, and the difference image signal The orthogonal transform coefficient is calculated by applying the orthogonal transform to the specified small area as a unit, and the orthogonal transform coefficient is calculated and expanded into bit planes. When performing processing for generating enhancement layer encoded data by encoding, (1) detection means for detecting an orthogonal transform coefficient having a value in the most significant bit plane, and (2) detection of the detection means And a replacement means for reducing the absolute value of the orthogonal transform coefficient by the most significant bit.
[0026]
When this configuration is adopted, the orthogonal transformation coefficient reduced by the most significant bit is set as a new processing target, and the ratio of the orthogonal transformation coefficient having a value in the most significant bit plane of the new processing target to the entire frame is defined. If the threshold value is smaller than the threshold value, there may be provided an instruction unit for instructing to repeat the process of reducing the absolute value of the orthogonal transform coefficient by the most significant bit.
[0027]
The image encoding method of the present invention realized by the operation of each of the above processing means can be realized by a computer program, and the computer program is provided by being recorded on a suitable recording medium such as a semiconductor memory. The present invention is realized by being provided via a network, installed on the occasion of carrying out the present invention, and operating on a control means such as a CPU.
[0028]
In the image encoding apparatus of the present invention configured as described above, a difference image signal between the original image signal and the base layer decoded image signal is input, and the orthogonal transformation is performed on the difference image signal in units of a prescribed small area. Calculates the orthogonal transform coefficients by applying the coefficients, expands the coefficients into bit planes, and encodes a predetermined number of planes from the most significant bit plane to the least significant bit plane to generate enhancement layer encoded data. At the time of processing, an orthogonal transform coefficient having a value in the most significant bit plane is detected, and the absolute value of the detected orthogonal transform coefficient is reduced by one most significant bit.
[0029]
In this way, in the present invention, when encoding the specified number of planes from the most significant bit plane to the least significant bit plane to generate the enhancement layer encoded data, usually, the most significant In consideration of the fact that there are few orthogonal transform coefficients having values in the bit plane, the absolute value of the orthogonal transform coefficient having a value in the most significant bit plane is reduced by the most significant bit, and the absolute value of the specified number of planes is reduced. Since the configuration for performing the encoding is employed, the encoding according to the characteristics of the residual signal can be performed with a simple configuration, and the quality of the decoded image can be improved.
[0030]
Further, in order to achieve the above object, an image encoding apparatus according to the present invention is used in hierarchical image encoding, in which a difference image signal between an original image signal and a base layer decoded image signal is input, and the difference is calculated. An orthogonal transformation coefficient is calculated by performing an orthogonal transformation on a specified small area as a unit for the image signal, and the orthogonal transformation coefficient is developed into a bit plane, and the specified plane is shifted from the highest bit plane to the lower bit plane. When performing processing for generating enhancement layer encoded data by encoding several times, (1) detecting means for detecting an orthogonal transform coefficient having a value in the most significant bit plane, and (2) detecting means Determining means for determining whether the absolute value of the orthogonal transform coefficient having a value in the most significant bit plane is reduced by the most significant bit according to the orthogonal transform coefficient detected by (3). Depending on the determination result, it constitutes the absolute value of the orthogonal transform coefficients having a value in the bit plane of the uppermost so and a replacement means for reducing by one bit most significant.
[0031]
When this configuration is adopted, the orthogonal transformation coefficient reduced by the most significant bit is set as a new processing target, and the ratio of the orthogonal transformation coefficient having a value in the most significant bit plane of the new processing target to the entire frame is defined. If the threshold value is smaller than the threshold value, there may be provided an instruction unit for instructing to repeat the process of reducing the absolute value of the orthogonal transform coefficient by the most significant bit.
[0032]
The image encoding method of the present invention realized by the operation of each of the above processing means can be realized by a computer program, and the computer program is provided by being recorded on a suitable recording medium such as a semiconductor memory. The present invention is realized by being provided via a network, installed on the occasion of carrying out the present invention, and operating on a control means such as a CPU.
[0033]
In the image encoding apparatus of the present invention configured as described above, a difference image signal between the original image signal and the base layer decoded image signal is input, and the orthogonal transformation is performed on the difference image signal in units of a prescribed small area. Calculates the orthogonal transform coefficients by applying the coefficients, expands the coefficients into bit planes, and encodes a predetermined number of planes from the most significant bit plane to the least significant bit plane to generate enhancement layer encoded data. At the time of processing, an orthogonal transform coefficient having a value in the most significant bit plane is detected, and an absolute value of the orthogonal transform coefficient having a value in the most significant bit plane is detected in accordance with the detected orthogonal transform coefficient. Is to be reduced by the most significant bit.
[0034]
For example, (a) when the ratio of the orthogonal transform coefficient having a value in the most significant bit plane to the entire frame is smaller than a prescribed threshold, it is determined that the value is reduced by the most significant bit, or (b) the most significant bit is determined. If the number of orthogonal transform coefficients having a value in the upper bit plane is larger than a prescribed threshold value is smaller than the prescribed threshold value, it is determined that the value is reduced by the most significant bit, or (c). When the value of the orthogonal transform coefficient having a value in the highest-order bit plane is larger than a specified threshold value and the number of quantization parameters smaller than the specified threshold value is smaller than the specified threshold value, It is determined that the size is reduced by the bit.
[0035]
If it is determined that the absolute value of the orthogonal transform coefficient having a value in the most significant bit plane is reduced by the most significant bit according to the determination result, the orthogonal transform having a value in the most significant bit plane is determined. Processing is performed so that the absolute value of the coefficient is reduced by the most significant bit.
[0036]
In this way, according to the present invention, when encoding the specified number of planes from the most significant bit plane to the least significant bit plane to generate enhancement layer encoded data, When it is determined that there are few orthogonal transform coefficients having a value or that the orthogonal transform coefficient having a value in the most significant bit plane is substantially small, the orthogonal transform coefficient having a value in the most significant bit plane is used. Is reduced by the most significant one bit, and encoding is performed for the specified number of planes. Therefore, encoding according to the characteristics of the residual signal is performed with a simple configuration. It is possible to improve the quality of the decoded image.
[0037]
BEST MODE FOR CARRYING OUT THE INVENTION
Hereinafter, the present invention will be described in detail according to embodiments.
[0038]
FIG. 1 shows an embodiment of the present invention.
[0039]
Here, in the figure, the same components as those shown in FIG. 12 are denoted by the same symbols.
[0040]
As shown in the figure, the second encoder 106 that encodes the enhancement layer configured according to the present invention newly includes a coefficient processing unit 401 when implementing a hierarchical encoding scheme in FGS encoding. It is characterized by the following.
[0041]
The coefficient processing unit 401 is a DCT coefficient to be encoded in the MSB plane (highest bit plane).
2^max≦ | coef | ≦ 2^{max + 1}-1
DCT coefficient coef
coef_cut = sign (coef) * (2^max-1)
Execute the process to replace with.
[0042]
Here, max = vop-max-level, and sign (x) is

And represents the sign of x.
[0043]
For example, when max = 6, the coefficient processing unit 401
64 = 2⁶≦ | coef | ≦ 2^{6 + 1}-1 = 127
DCT coefficient coef in the range of
coef_cut = 2⁶-1 = 63
Will be executed.
[0044]
That is, the coefficient processing unit 401 performs processing to reduce the absolute value of the DCT coefficient to be encoded by the most significant bit. Hereinafter, the processing performed by the coefficient processing unit 401 will be referred to as “peak processing (DCT coefficient peak processing)”.
[0045]
Next, the encoding process executed by the second encoder 106 configured as shown in FIG. 1 will be described in detail according to the processing flow of FIG.
[0046]
As shown in the processing flow of FIG. 2, the second encoder 106 first inputs a differential image signal to be encoded in step 10, and then, in step 11, inputs the differential image signal , DCT coefficients are calculated by performing discrete cosine transform for each macroblock.
[0047]
Subsequently, in step 12, a bit shift process is performed on the calculated DCT coefficient for each macroblock.
[0048]
Subsequently, in step 13, the number of bit planes (max described above) is determined by searching for the DCT coefficient having the largest absolute value among the DCT coefficients of the entire frame. For example, when the absolute value of the DCT coefficient of 65 is the maximum value, it is determined that the number of bit planes is max = 6.
[0049]
Subsequently, at step 14, it is determined whether or not the processing has been completed for all the macroblocks of the frame, and when it is determined that unprocessed macroblocks are left, the process proceeds to step 15, where unprocessed macroblocks are processed. Select one macroblock.
[0050]
Subsequently, in step 16, the above-described peaking process is performed on the DCT coefficient having a value in the most significant bit plane for the selected macroblock, and the process returns to step 14. For example, a DCT coefficient having an absolute value of 65 is subjected to a censoring process for replacing the DCT coefficient so that its absolute value becomes 63, and then the process returns to step 14.
[0051]
As described above, when it is determined in step 14 that the processing has been completed for all the macroblocks of the frame while repeating the processing of step 14 to step 16, the process proceeds to step 17, and the processing is performed from the most significant bit plane. The variable-length coding is performed for the lower bit planes by the number of bit planes defined by the FGS, and the process ends. Specifically, the variable length coding is performed on the upper four planes, and the processing is terminated.
[0052]
Next, the effectiveness of the encoding processing of the present invention configured as described above will be verified using a specific calculation example.
[0053]
When the discrete cosine transform is performed on the image difference values of 8 × 8 pixels shown in FIG. 3A, the DCT coefficients shown in FIG. 3B are obtained.
[0054]
When the encoding process according to the related art is followed, variable-length encoding is performed on the upper four planes of the DCT coefficients shown in FIG. 3B, and the resulting data is decoded to obtain a decoding difference value shown in FIG. You will get.
[0055]
On the other hand, in the encoding process of the present invention, the DCT coefficients shown in FIG. 3B are subjected to the above-described peaking process to be converted into the DCT coefficients shown in FIG. Variable-length coding is performed on the upper four planes of the coefficients, and decoding is performed to obtain a decoded difference value shown in FIG. That is, according to the above-described peaking process, the variable-length coding is performed while converting the DCT coefficient having the absolute value of 65 into the one having the absolute value of 63.
[0056]
From the image difference value shown in FIG. 3A and the decoding difference value shown in FIG. 3D, when the encoding process of the related art is followed, the maximum error is 12, and the total error is 208. On the other hand, as can be seen from the image difference value shown in FIG. 3A and the decoding difference value shown in FIG. 3E, the maximum error is 7 and the error The total is 133.
[0057]
As described above, according to the encoding processing of the present invention, it is possible to improve the quality of a decoded image.
[0058]
When the coefficient processing unit 401 follows the processing flow of FIG. 2, the peak processing is performed once on the most significant bit plane. However, the DCT coefficient reduced by the most significant one bit is set as a new processing target. According to the number of DCT coefficients having a value in the most significant bit plane of the new processing target, it is also possible to perform processing to further continue the peaking processing.
[0059]
The encoding process executed by the second encoder 106 that realizes this configuration will be described in detail according to the process flow illustrated in FIG.
[0060]
When the second encoder 106 follows the processing flow of FIG. 4, first, in step 20, the second encoder 106 inputs a differential image signal to be encoded, and then, in step 21, the differential image signal is input. , A DCT coefficient is calculated by performing discrete cosine transform for each macroblock.
[0061]
Subsequently, in step 22, a bit shift process is performed on the calculated DCT coefficient for each macroblock, and in the following step 23, the DCT coefficient of the entire frame having the largest absolute value is searched to obtain Determine the number of planes.
[0062]
Subsequently, in step 24, it is determined whether or not the processing has been completed for all the macroblocks of the frame, and when it is determined that the unprocessed macroblock remains, the process proceeds to step 25, where the unprocessed macroblock is processed. One macro block is selected, and in the following step 26, the selected macro block is subjected to a peaking process on DCT coefficients having values in the most significant bit plane, and the process returns to step 24.
[0063]
As described above, when it is determined in step 24 that the processing has been completed for all the macroblocks of the frame while repeating the processing of step 24 to step 26, the process proceeds to step 27, where the DCT that has been subjected to the peaking processing is executed. The coefficient is set as a new DCT coefficient of the frame, and in the following step 28, the number of bit planes is determined by searching for the DCT coefficient of the entire frame having the largest absolute value.
[0064]
Subsequently, in step 29, it is determined whether or not the ratio of the DCT coefficient having a value in the most significant bit plane to the entire frame is larger than a prescribed threshold value. On the other hand, the process returns to step 24 to continue the peaking process.
[0065]
On the other hand, when it is determined in step 29 that the ratio of the DCT coefficient having a value in the most significant bit plane to the entire frame is larger than a prescribed threshold, it is determined that it is not appropriate to continue the peaking process, and Proceeding to 30, the variable length encoding is performed for the number of bit planes defined by the FGS from the most significant bit plane to the least significant bit plane, and the process ends.
[0066]
In this way, when the coefficient processing unit 401 follows the processing flow of FIG. 4, the DCT coefficient reduced by the most significant one bit is set as a new processing target, and the DCT coefficient is added to the highest-order bit plane of the new processing target. According to the number of DCT coefficients having a value, processing is performed so as to continue the peak processing.
[0067]
FIG. 5 shows another embodiment of the present invention.
[0068]
Here, in the figure, the same components as those shown in FIG. 1 are indicated by the same symbols.
[0069]
According to this embodiment, the second encoder 106 that encodes the enhancement layer configured according to the present invention newly adds a coefficient count / determination unit 402 when implementing the hierarchical encoding scheme in the FGS encoding. It is characterized by having.
[0070]
The coefficient count / determination unit 402 calculates a ratio p of the DCT coefficients included in the most significant bit plane to the entire frame, and performs a process of determining whether to perform the peaking process according to the ratio p.
[0071]
Next, the encoding process executed by the second encoder 106 configured as shown in FIG. 5 will be described in detail according to the processing flow of FIG.
[0072]
When configured as shown in FIG. 5, the second encoder 106 first inputs a differential image signal to be encoded in step 40 as shown in the processing flow of FIG. Subsequently, in step 41, a DCT coefficient is calculated by applying a discrete cosine transform to the difference image signal for each macroblock.
[0073]
Subsequently, in step 42, a bit shift process is performed on the calculated DCT coefficient for each macroblock, and in a succeeding step 43, the DCT coefficient having the largest absolute value is searched for among the DCT coefficients of the entire frame, thereby obtaining the bit value. Determine the number of planes.
[0074]
Subsequently, in step 44, it is determined whether or not the ratio of the DCT coefficient having a value in the most significant bit plane to the entire frame is larger than a prescribed threshold value. Is determined not to be appropriate, the process proceeds to step 45, variable-length encoding is performed from the most significant bit plane to the least significant bit planes by the number of bit planes defined by FGS, and the process is terminated.
[0075]
That is, if the ratio of the DCT coefficient having a value in the most significant bit plane to the entire frame is larger than a specified threshold, the peak processing is performed from the most significant bit plane to the least significant bit plane without performing the peaking process. The variable length coding is performed for the number of bit planes defined by the FGS.
[0076]
On the other hand, when it is determined in step 44 that the ratio of the DCT coefficient having the value in the most significant bit plane to the entire frame is not larger than the specified threshold, the process proceeds to step 46, in which the processing is performed for all the macroblocks in the frame. When it is determined whether or not the processing has been completed and it is determined that an unprocessed macroblock remains, the process proceeds to step 47, and one unprocessed macroblock is selected.
[0077]
Subsequently, in step 48, the selected macroblock is subjected to a peaking process on DCT coefficients having values in the most significant bit plane.
[0078]
In this way, when it is determined in step 46 that the processing has been completed for all the macroblocks of the frame while repeating the processing of step 46 to step 48, the process proceeds to step 45, and The variable length coding is performed for the lower bit planes by the number of bit planes defined by the FGS, and the process ends.
[0079]
That is, when the ratio of the DCT coefficient having a value in the most significant bit plane to the entire frame is smaller than a predetermined threshold, the DCT coefficient is subjected to a peaking process, and the DCT coefficient subjected to the peaking process is encoded. The variable length coding is performed for the number of bit planes defined by the FGS from the most significant bit plane to the least significant bit plane as an object to be converted.
[0080]
In this way, when configured as shown in FIG. 5, the second encoder 106 obtains the ratio p of the DCT coefficients included in the most significant bit plane to the entire frame, and sets the ratio p to a predetermined value. The threshold value Th1 is compared with the threshold value Th, and if p ≦ Th1, processing is performed to perform a peaking process.
[0081]
According to this configuration, it is possible to avoid the problem that the accuracy of the decoded residual signal is deteriorated, which occurs when the ratio of the DCT coefficient to be subjected to the peaking process is large.
[0082]
When the coefficient count / determination unit 402 follows the processing flow of FIG. 6, it does not consider the distribution of DCT coefficients having values in the most significant bit plane, but considers this distribution and determines whether to perform the peaking process. It is also possible to adopt a configuration in which it is determined that
[0083]
The encoding process executed by the second encoder 106 that realizes this configuration will be described in detail with reference to the process flow illustrated in FIG.
[0084]
When the second encoder 106 follows the processing flow of FIG. 7, first, in step 50, the second encoder 106 inputs a differential image signal to be encoded, and subsequently, in step 51, the differential image signal is input. , A DCT coefficient is calculated by performing discrete cosine transform for each macroblock.
[0085]
Subsequently, in step 52, a bit shift process is performed on the calculated DCT coefficient for each macroblock, and in the following step 53, the DCT coefficient having the largest absolute value among the DCT coefficients of the entire frame is searched for, thereby obtaining the bit value. Determine the number of planes.
[0086]
Subsequently, in step 54, the number of DCT coefficients having a large value among the DCT coefficients having values in the most significant bit plane is determined.
[0087]
For example, for a certain threshold Th2,
2^max≦ | coef | ≦ 2^max+ Th2
A set of coefs satisfying the condition is expressed as {COEF2},
2^max+ Th2 <| coef | <2^{max + 1}
Is expressed as {COEF3}, the number N (COEF3) of coefs belonging to the {COEF3} is given by a threshold Th3.
N (COEF3) ≦ Th3
Is determined, the number of DCT coefficients having a large value in the most significant bit plane is determined.
[0088]
Subsequently, in step 55, it is determined whether or not the result of the determination processing in step 54 that the number of DCT coefficients having a value in the most significant bit plane having a large value is large is large. If it is determined that there are many, the process proceeds to step 56 where variable length coding is performed from the most significant bit plane to the least significant bit planes by the number of bit planes specified by FGS, and the processing is performed. finish.
[0089]
In other words, when the number of DCT coefficients having a large value in the most significant bit plane is large, the FGS is performed from the most significant bit plane to the least significant bit plane without performing the peaking process. The variable length coding is performed for the specified number of bit planes.
[0090]
On the other hand, if it is determined in step 55 that the number of DCT coefficients having a value in the most significant bit plane having a large value is small, the process proceeds to step 57 to determine whether the processing has been completed for all macroblocks of the frame. When it is determined whether or not there is an unprocessed macroblock remaining, the process proceeds to step 58, and one unprocessed macroblock is selected.
[0091]
Subsequently, in step 59, the selected macroblock is subjected to a peaking process on DCT coefficients having values in the most significant bit plane.
[0092]
As described above, when it is determined in step 57 that the processing has been completed for all the macroblocks of the frame when the processing of steps 57 to 59 is repeated, the process proceeds to step 56, where the most significant bit plane The variable length coding is performed for the lower bit planes by the number of bit planes defined by the FGS, and the process ends.
[0093]
That is, when the number of DCT coefficients having a large value among the DCT coefficients having values in the most significant bit plane is small, the DCT coefficients are subjected to a peaking process, and the DCT coefficients subjected to the peaking process are subjected to encoding. The variable length coding is performed from the most significant bit plane to the least significant bit plane by the number of bit planes defined by the FGS.
[0094]
In this way, when following the processing flow of FIG. 7, the coefficient count / determination unit 402 sets the number N (COEF3) of DCT coefficients having values in the most significant bit plane having a large value and The threshold value Th3 is compared with the threshold value Th, and if N (COEF3) ≦ Th3, processing is performed to perform a peaking process.
[0095]
According to this configuration, it is possible to avoid the problem that the precision of the decoded residual signal deteriorates, which occurs when the value of the DCT coefficient changes due to the peaking process is large.
[0096]
FIG. 8 illustrates another embodiment of the present invention.
[0097]
Here, in the figure, the same components as those shown in FIG. 5 are denoted by the same symbols.
[0098]
In accordance with this embodiment, the coefficient count / determination unit 402 receives the encoding parameter from the control unit 206 (shown in FIG. 11) constituting the first encoder 101 via the signal line 403. And performing a process of determining whether or not to perform the peaking process using the encoding parameter.
[0099]
For example, taking advantage of the fact that the quantization parameter in the first encoder 101 is likely to be large in a macroblock corresponding to a complex area of the original image, the coefficient counting / determining unit 402 uses the first encoder 101 Receiving the quantization parameter QP from the control unit 206 constituting
N (COEF4) ≦ Th5
Is determined, and if this condition is satisfied, it is determined to execute the peaking process.
[0100]
Here, the set {COEF4} is defined as an element x of the above set {COEF3} that satisfies QPx ≦ Th4. Here, QPx is a quantization parameter in the base layer of the macroblock including the DCT coefficient x, and N {COEF4} is the number of DCT coefficients belonging to the set {COEF4}.
[0101]
Next, the encoding process executed by the second encoder 106 configured as shown in FIG. 8 will be described in detail according to the processing flow of FIG.
[0102]
In the case where the second encoder 106 is configured as shown in FIG. 8, as shown in the processing flow of FIG. 9, first, in step 60, when the differential image signal to be encoded is input, Subsequently, in step 61, a DCT coefficient is calculated by performing a discrete cosine transform on the difference image signal for each macroblock.
[0103]
Subsequently, in step 62, a bit shift process is performed on the calculated DCT coefficient for each macroblock, and in a succeeding step 63, the DCT coefficient of the entire frame having the largest absolute value is searched to obtain Determine the number of planes.
[0104]
Subsequently, at step 64, among the DCT coefficients having values in the most significant bit plane,
N (COEF4) ≦ Th5
, The number of quantization parameters having small values is determined.
[0105]
Subsequently, in step 65, it is determined whether or not the condition of N (COEF4) ≦ Th5 is satisfied by the determination processing in step 64. If a determination result that this condition is not satisfied is obtained, step 66 is performed. The variable length coding is performed for the number of bit planes defined by the FGS from the most significant bit plane to the least significant bit plane, and the process is terminated.
[0106]
That is, when the condition of N (COEF4) ≦ Th5 is not satisfied, and therefore, when the number of N (COEF4) is large, the FGS is performed from the uppermost bit plane to the lower bit plane without performing the peaking process. The variable-length coding is performed for the number of bit planes defined by the above.
[0107]
On the other hand, when it is determined in step 65 that the condition of N (COEF4) ≦ Th5 is satisfied, the process proceeds to step 67, in which it is determined whether or not processing has been completed for all macroblocks of the frame. When it is determined that the remaining macroblock remains, the process proceeds to step 68, where one unprocessed macroblock is selected.
[0108]
Subsequently, in step 69, the selected macroblock is subjected to a peaking process on DCT coefficients having values in the most significant bit plane.
[0109]
As described above, when it is determined in step 67 that the processing has been completed for all the macroblocks of the frame while repeating the processing of steps 67 to 69, the process proceeds to step 66, where The variable length coding is performed for the lower bit planes by the number of bit planes defined by the FGS, and the process ends.
[0110]
That is, when the condition of N (COEF4) ≦ Th5 is satisfied, that is, when the number of N (COEF4) is small, the DCT coefficient is subjected to a peaking process, and the DCT coefficient subjected to the peaking process is encoded. As an object, variable-length coding is performed from the most significant bit plane to the least significant bit plane by the number of bit planes specified by FGS.
[0111]
In this way, when following the processing flow of FIG. 9, the coefficient count / determination unit 402 determines, among the elements x of the above-mentioned set {COEF3}, that of the set {COEF4} defined as satisfying QPx ≦ Th4. The number N (COEF4) is compared with a preset threshold value Th5, and if N (COEF4) ≦ Th5, processing is performed to perform a peaking process.
[0112]
According to this configuration, a DCT coefficient belonging to such a region is excluded from the above-described determination of N (COEF3) ≦ Th3 by utilizing a visual characteristic that an error of a decoded image is less noticeable in a complicated region. is there.
[0113]
As a result, it is possible to reliably avoid the problem that the accuracy of the decoded residual signal deteriorates, which occurs when the value of the DCT coefficient changes due to the peaking process is large.
[0114]
Although the present invention has been described with reference to the illustrated embodiment, the present invention is not limited thereto. For example, although not described in the embodiment shown in FIGS. 5 and 8, in the embodiment shown in FIGS. 5 and 8, a DCT coefficient reduced by the most significant bit is set as a new processing target. The processing described in the processing flow of FIG. 4 in which the peak processing is further continued in accordance with the number of DCT coefficients having values in the most significant bit plane of the new processing target is used. It may be.
[0115]
【The invention's effect】
As described above, according to the present invention, when generating the enhancement layer encoded data by encoding the specified number of planes from the most significant bit plane to the least significant bit plane, the normal In consideration of the fact that the number of orthogonal transform coefficients having values in the bit planes is small, the absolute value of the orthogonal transform coefficients having values in the most significant bit plane is reduced by the most significant bit, and the absolute number of , The encoding according to the characteristics of the residual signal can be performed with a simple configuration to improve the quality of the decoded image.
[0116]
In addition, in the present invention, when the enhancement layer coded data is generated by coding the specified number of planes from the most significant bit plane to the least significant bit plane, the orthogonal bit plane having a value in the most significant bit plane is generated. When judging that the number of transform coefficients is small or that the number of orthogonal transform coefficients having a value in the most significant bit plane is substantially small, the absolute value of the orthogonal transform coefficient having a value in the most significant bit plane Is reduced by the most significant bit, and encoding is performed for the specified number of planes. Therefore, encoding according to the characteristics of the residual signal is performed with a simple configuration, and The quality can be improved.
[Brief description of the drawings]
FIG. 1 is an embodiment of the present invention.
FIG. 2 is an example of a processing flow executed by the present invention.
FIG. 3 is a diagram for explaining the effectiveness of the encoding process of the present invention.
FIG. 4 is an example of a processing flow executed by the present invention.
FIG. 5 is another embodiment of the present invention.
FIG. 6 is an example of a processing flow executed by the present invention.
FIG. 7 is an example of a processing flow executed by the present invention.
FIG. 8 is another embodiment of the present invention.
FIG. 9 is an example of a processing flow executed by the present invention.
FIG. 10 is an explanatory diagram of a hierarchical encoding process.
FIG. 11 is an explanatory diagram of a hierarchical encoding process.
FIG. 12 is an explanatory diagram of a hierarchical encoding process.
FIG. 13 is an explanatory diagram of DCT coefficients.
FIG. 14 is an explanatory diagram of an FGS encoding process.
[Explanation of symbols]
105 difference image signal
106 second encoder
108 Enhancement layer encoded data
301 Discrete Cosine Transform Processing Unit
302 weighting unit
303 Maximum value search unit
304 variable length coding unit
401 coefficient processing unit

Claims (14)

Used in hierarchical image coding, an orthogonal transformation is performed by inputting a difference image signal between an original image signal and a base layer decoded image signal, and performing orthogonal transformation on the difference image signal in units of a prescribed small area. In an image coding apparatus that calculates coefficients, expands them into bit planes, and codes the specified number of planes from the most significant bit plane to the lower bit planes to generate enhancement layer encoded data. ,
Means for detecting an orthogonal transform coefficient having a value in the most significant bit plane;
Means for reducing the absolute value of the detected orthogonal transform coefficient by one most significant bit.
An image encoding device characterized by the following. Used in hierarchical image coding, an orthogonal transformation is performed by inputting a difference image signal between an original image signal and a base layer decoded image signal, and performing orthogonal transformation on the difference image signal in units of a prescribed small area. In an image coding apparatus that calculates coefficients, expands them into bit planes, and codes the specified number of planes from the most significant bit plane to the lower bit planes to generate enhancement layer encoded data. ,
Means for detecting an orthogonal transform coefficient having a value in the most significant bit plane;
Means for determining whether to reduce the absolute value of the orthogonal transform coefficient having a value in the most significant bit plane by the most significant bit according to the detected orthogonal transform coefficient,
Means for reducing the absolute value of the orthogonal transform coefficient having a value in the most significant bit plane by one most significant bit according to the determination result.
An image encoding device characterized by the following. The image encoding device according to claim 2,
The determining means determines an absolute value of the orthogonal transform coefficient having a value in the most significant bit plane when a ratio of the orthogonal transform coefficient having a value in the most significant bit plane to the entire frame is smaller than a predetermined threshold. Is determined to be smaller by one most significant bit.
An image encoding device characterized by the following. The image encoding device according to claim 2,
The determining means sets the value to the most significant bit plane when the number of orthogonal transform coefficients having a value in the most significant bit plane is larger than a prescribed threshold is less than a prescribed threshold. To determine that the absolute value of the orthogonal transform coefficient to have is reduced by the most significant bit,
An image encoding device characterized by the following. The image encoding device according to claim 2,
The determining means includes, among the orthogonal transform coefficients having values in the most significant bit plane that are larger than a specified threshold, the number of the quantization parameters whose values are smaller than the specified threshold is smaller than the specified threshold. When the number is small, it is determined that the absolute value of the orthogonal transform coefficient having a value in the most significant bit plane is reduced by the most significant bit.
An image encoding device characterized by the following. The image encoding device according to any one of claims 1 to 5,
If the orthogonal transform coefficient reduced by the most significant one bit is set as a new processing target and the ratio of the orthogonal conversion coefficient having a value in the most significant bit plane of the new processing target to the entire frame is smaller than a prescribed threshold value, And means for instructing to repeat the process of reducing the absolute value of the orthogonal transform coefficient by the most significant bit.
An image encoding device characterized by the following. Used in hierarchical image coding, an orthogonal transformation is performed by inputting a difference image signal between an original image signal and a base layer decoded image signal, and performing orthogonal transformation on the difference image signal in units of a prescribed small area. A coefficient is calculated, developed on a bit plane, and encoded by a specified number of planes from the most significant bit plane to the least significant bit plane, thereby generating an enhancement layer encoded data. ,
Detecting an orthogonal transform coefficient having a value in the most significant bit plane;
Reducing the absolute value of the detected orthogonal transform coefficient by one most significant bit.
Characteristic image coding method. Used in hierarchical image coding, an orthogonal transformation is performed by inputting a difference image signal between an original image signal and a base layer decoded image signal, and performing orthogonal transformation on the difference image signal in units of a prescribed small area. A coefficient is calculated, developed on a bit plane, and encoded by a specified number of planes from the most significant bit plane to the least significant bit plane, thereby generating an enhancement layer encoded data. ,
Detecting an orthogonal transform coefficient having a value in the most significant bit plane;
A step of determining whether to reduce the absolute value of the orthogonal transform coefficient having a value in the most significant bit plane by the most significant bit according to the detected orthogonal transform coefficient;
Reducing the absolute value of the orthogonal transform coefficient having a value in the most significant bit plane by one most significant bit according to the determination result.
Characteristic image coding method. The image encoding method according to claim 8,
In the determining step, when the ratio of the orthogonal transform coefficient having a value in the most significant bit plane to the entire frame is smaller than a predetermined threshold, the absolute value of the orthogonal transform coefficient having a value in the most significant bit plane is determined. Is determined to be smaller by one most significant bit.
Characteristic image coding method. The image encoding method according to claim 8,
In the determining step, when the value of the orthogonal transform coefficient having a value in the most significant bit plane is larger than a prescribed threshold value and the number is smaller than a prescribed threshold value, the value is assigned to the most significant bit plane. To determine that the absolute value of the orthogonal transform coefficient to have is reduced by the most significant bit,
Characteristic image coding method. The image encoding method according to claim 8,
In the determining step, the number of the orthogonal transform coefficients having values in the most significant bit planes that are larger than a specified threshold is smaller than the specified threshold for the number of quantization parameters whose values are smaller than the specified threshold. When the number is small, it is determined that the absolute value of the orthogonal transform coefficient having a value in the most significant bit plane is reduced by the most significant bit.
Characteristic image coding method. The image encoding method according to any one of claims 7 to 11,
If the orthogonal transform coefficient reduced by the most significant one bit is set as a new processing target and the ratio of the orthogonal conversion coefficient having a value in the most significant bit plane of the new processing target to the entire frame is smaller than a prescribed threshold value, Including a process of instructing to repeat the process of reducing the absolute value of the orthogonal transform coefficient by the most significant bit.
Characteristic image coding method. An image encoding program for causing a computer to execute processing used for implementing the image encoding method according to claim 7. A recording medium storing an image encoding program for causing a computer to execute a process used for implementing the image encoding method according to any one of claims 7 to 12.

Images