JP3819638B2 - Variable length coding device - Google Patents

Description

この時、いかに重み付け係数ｗ（θi ）を求めるかが問題となる。
【０１５３】
情報源θi をＱ（Ｘ）で符号化した時の理想符号長Ｌ（Ｘ| θi ）は、

となる。各情報源に対して平均的に理想符号長Ｌ（Ｘ| θi ）を最小化するには

【０１５４】
この関数が最小になるのは、Ｕ（Ｘ）＝Ｑ（Ｘ）の時で、
ｗ（θ1 ）＝…＝ｗ（θn ）＝１／ｎ
となる。
【０１５５】
別の確率モデルＱ（Ｘ）の設計方法として、最悪の情報源を想定して確率モデルＱ（Ｘ）を設計する手法を示す。
【０１５６】
情報源θi をＱ（Ｘ）で符号化した時の冗長度Ｒ（Ｘ| θi ）は、

となる。この冗長度の各情報源に対する重みづけ平均Ｓ（Ｘ）は、

となって、事象Ｘと事象θの相互情報量を示す関数になっている。この関数の最大値を求める方法は、
S.Arimoto,"An algorithm for computing the capacity of arbitrary discretememoryless channels,"IEEE Trans.Inform.Theory,Vol.IT-18,pp.14-20,1972.
R.E.Blahut,"Computation of channel capacity and rate-distortion functions,"IEEE Trans.Inform.Theory,Vol.IT-18,pp.460-473,1972.
に示されるArimoto-Blahutのアルゴリズム等が知られており、これらのアルゴリズムによってＳ（Ｘ）を最大化するつまり最悪の場合のｗ（θi ）（i=1,…n ）を求めることができる。
【０１５７】
確率モデルＱ（Ｘ）の設計方法として、理想符号長を平均的に最小にする方法と、平均的に最悪を想定する方法を示したが、２つの方法を組み合わせて、情報源をグループ化して前者の方法で確率モデルをいくつか作成し、その確率モデルに対して後者の方法を用いるなどの方法を適用してもよい。
【０１５８】
図５０に示す符号選択部２１では、確率モデル作成部２３で作成した確率モデルＱ（Ｘ）を、情報シンボルＸを確率の高い順にソーティングしたＱ（Ｙ）を作成し、一方、符号構成部２２で作成されたリバーシブル符号の符号長を短い順にソーティングしたものＦ（Ｚ）を作成し、

を計算して最も値が小さいものを選択し、情報シンボルと符号語を対応させて符号語テーブルを作成する。
【０１５９】
符号語構成部２２で構成する符号は、例えば特願平７−８９７７２および特願平７−２６０３８３並びに第１および第３実施形態で示されている符号語の重みを用いた順方向にも逆方向にも復号可能な可変長符号が用いられる。
【０１６０】
（第７実施形態）
次に、本発明の第７実施形態を説明する。図５１は、本発明の第７実施形態による可変長符号化／復号化装置が組み込まれた動画像符号化／復号化装置の概念図である。動画像符号化器７０９では、情報源符号化器７０２で符号化されたデータは、動画像多重化部７０３で可変長符号化、多重化等がおこなわれ、伝送バッファ７０４で平滑化されたのちに、符号化データとして、伝送系または蓄積系７０５に送り出される。符号化制御部７０１は、伝送バッファ７０４のバッファ量を考慮して、情報源符号化器７０２、動画像多重化部７０３の制御を行なう。一方、動画像復号化器７１０では、伝送系または蓄積系７０５からの符号化データを受信バッファ７０６にためてから、動画像多重化分離部７０７で、符号化データの多重化分離、可変長復号化が行なわれ、情報源復号化器７０８に送られ、最終的に動画像が復号化される。
【０１６１】
第７実施形態において追加されたシンタクスについて説明する。図５３は、第７実施形態における動画像多重化部７０３および、動画像多重化分離部７０７での動画像符号化方式のシンタクスを示している。符号化データは、上位階層と下位階層に階層化し、それぞれを同期符号ＲＭとＭＭによって、同期区間を設定している。また、下位階層のＳＴは、図６２で示すような逆方向から復号可能なスタッフィング符号である。フレーム内符号化フレームは、図６３のように、マクロブロックのモード情報の一部とＩＮＴＲＡ ＤＣを上位階層とし、モード情報の一部とＡＣ−ＤＣＴ係数情報を下位階層として、ＡＣ−ＤＣＴ係数情報にリバーシブル符号を適用している。フレーム間符号化フレームは、図６４のように、マクロブロックのモード情報の一部と動きベクトル情報を上位階層とし、モード情報の一部とＩＮＴＲＡ ＤＣとＡＣ−ＤＣＴ係数情報を下位階層として、ＡＣ−ＤＣＴ係数情報にリバーシブル符号を適用している。
【０１６２】
図５３に示す上位階層と下位階層の符号化および復号化は、図５２に示す動画像多重化部とにより行なわれている。これらの構成は、第５実施形態に係る符号化装置／復号化装置の構成と略同一である。すなわち、図５１の動画像多重化部７０３および動画像多重化分離部７０７の基本構成は、図５３（ａ）（ｂ）に示す通りであるが、上位階層可変長符号化器８０１、下位階層可変長符号化器８０３および上位階層可変長復号化器８０５、下位階層可変長復号化器８０６の構成が第２実施形態とは異なっている。すなわち、情報源符号化器７０２で符号化されたデータのうち、例えば図５３（ａ）のシンタクスで示された上位階層のデータが上位階層可変長符号化器８０１で可変長符号化され、多重化部８０３に送られる。また、情報源符号化器７０２で符号化されたデータのうち、図５３（ｂ）のシンタクスで示された下位階層のデータが可変長符号化器８０３で可変長符号化され、多重化部８０３に送られる。多重化部８０３では、上位階層の符号化データおよび下位階層の符号化データを多重化し、送信バッファ７０４に送る。
【０１６３】
そして、この第７実施形態では上位階層可変長復号化器８０５および下位階層可変長復号化器８０６内の図３２中に示した復号値判定部１１０では、順方向復号化器１０８によって得られた順方向復号結果から復号値を判定し、最終的な復号結果を出力する。各階層の順方向復号化器１０８および逆方向復号化器１０９での誤り判定は、符号語として存在しないビットパターンが出現した場合、およびチェックビット等で誤りを検出した場合には、検出位置はその場所とし、前記の判定方法で誤りが検出されない場合で、復号したビット数が同期区間内の符号化データのビット数と一致しない場合は、復号化の最処理位置を誤り検出位置とする。
【０１６４】
（第８実施形態）
図５４，図５５は、本発明の第８実施形態に係る符号化／復号化装置における情報源符号化器７０２，情報源復号化器７０８の一例をそれぞれ示している。まず、図５４において、入力画像信号は、ブロック化回路４０１でマクロブロックに分割される。マクロブロックに分割された入力画像信号は、減算器４０２に入力され、予測画像信号との差分がとられて、予測残差信号が生成される。この予測残差信号と、ブロック化回路４０１からの入力画像信号のいずれか一方を、モード選択スイッチ４０３により選択し、ＤＣＴ（離散コサイン変換）回路４０４により離散コサイン変換される。ＤＣＴ回路で得られたＤＣＴ係数データは、量子化回路４０５で量子化される。量子化されたデータは、２分岐され、一方はＤＣＴ係数情報として、動画像多重化器７０３におくられる。もう一方は、逆量子化回路４０６およびＩＤＣＴ（逆離散コサイン変換）回路４０７により量子化回路４０５およびＤＣＴ回路４０６の処理と逆の処理が順次行なわれた後、加算器４０８でスイッチ４１０を介して入力される予測画像信号と加算されることにより、局部復号信号が生成される。この局部復号信号は、フレームメモリに蓄えられ、動き補償回路４１１に入力される。動き補償回路Ａ０１０では、入力画像信号と局部復号信号との間で動き補償を行ない、動きベクトルを求め、予測画像信号を生成する。このとき求めめられた動きベクトル情報は、動画像多重化器７０３に送られる。モード選択回路４１２では、動き補償回路４１０からの情報に基づいて、各マクロブロックをどのような予測方式で符号化するかを決定し、その結果が量子化パラメータ等ともにモード情報として動画像多重化器７０３に送られる。なお、図５４において、出力される符号化データはＤＣＴ符号化係数，動きベクトル情報，モード情報の３つが１つに纏められたデータである。
【０１６５】
次に、図５４により符号化されたデータを復号化する情報源復号化器７０８の構成について図５５を参照しながら説明する。図５５は、図５４に対応する情報源復号化器７０８の一例を示している。図５５において、情報源復号化器７０８は、動画像多重化分離回路７０７で分離したモード情報、動きベクトル情報、ＤＣＴ係数情報等が入力される。
【０１６６】
モード情報が入力されるモード判定回路５０４では、モード情報がＩＮＴＲＡならば、モード切替スイッチ５０５をオフに選択して、ＤＣＴ係数情報を、逆量子化回路５０１で逆量子化し、ＩＤＣＴ回路５０２で逆離散コサイン変換処理を行なうことにより、再生画像信号を生成させる。この再生画像信号は、フレームメモリ５０６に参照画像として蓄積される一方、再生画像信号として出力する。モード情報かＩＮＴＥＲならば、モード切替スイッチ５０５をオフに選択して、ＤＣＴ係数情報を、逆量子化回路５０１で逆量子化し、ＩＤＣＴ回路５０２で逆離散コサイン変換処理をおこない、動きベクトル情報に基づいて、フレームメモリ５０４において参照画像を動き補償し、加算器５０３で足しあわせて、再生画像信号を生成させる。この再生画像信号は、フレームメモリ５０４に参照画像として蓄積される一方、再生画像信号として出力する。
【０１６７】
（第９実施形態）
次に、本発明の応用例として、本発明の可変長符号化／復号化装置が組み込まれた第９実施形態に係る動画像符号化／復号化システムを適用した画像送受信装置の実施形態を図５６を用いて説明する。パーソナルコンピュータ（ＰＣ）１００１に備え付けられたカメラ１００２より入力された画像信号は、ＰＣ１００１に組み込まれた動画像符号化器７０９によって符号化される。動画像符号化器７０９から出力される符号化データは、他の音声やデータの情報と多重化された後、無線機１００３より無線で送信され、他の無線機１００４によって受信される。無線機１００４で受信された信号は、画像信号の符号化データおよび音声データに分解される。これらのうち、画像信号の符号化データはワークステーション（ＥＷＳ）１００５に組み込まれた動画像復号化器７１０によって復号され、ＥＷＳ１００５のディスプレイに表示される。
【０１６８】
一方、ＥＷＳ１００５に備え付けられたカメラ１００６より入力された画像信号は、ＥＷＳ１００５に組み込まれた動画像符号化器７０９を用いて上記と同様に符号化される。符号化データは、他の音声やデータの情報と多重化され、無線機１００４により無線で送信され、無線機１００３によって受信される。無線機１００３によって受信された信号は、画像信号の符号化データおよび音声やデータの情報に分解される。これらのうち、画像信号の符号化データはＰＣ１００１に組み込まれた動画像復号化器７１０によって復号され、ＰＣ１００１のディスプレイに表示される。
【０１６９】
（第１０実施形態）
なお、上述した第１ないし第９実施形態に係る動画像符号化／復号化装置は、ハードウェアとしての構成として発明を捉えていたが、本発明はこれを発展させて上記各々の実施形態に係る符号化／復号化装置において用いられるデータまたはプログラムを記録する記録媒体としても捉えることができる。以下、第１０実施形態に係るデータまたはプログラムを記録した記録媒体について説明する。
【０１７０】
図５１に示す情報源符号化器７０２では、量子化語の８×８のＤＣＴ係数のブロックについてブロックないのスキャンを行なって、ＬＡＳＴ（０：ブロックの最後でない非零係数，１：ブロックの最後の非零係数），ＲＵＮ（非零係数までの零ランの数）及びＬＥＶＥＬ（係数の量子化値）を求め、動画像多重化部７０３に送る。
【０１７１】
動画像多重化部７０３内の下位階層可変長符号化器８０２は、図６０に示すＩＮＴＲＡ（フレーム内符号化）の非ＬＡＳＴ係数のＲＵＮ，ＬＥＶＥＬにリバーシブル符号の符号語（ＶＬＣ＿ＣＯＤＥ）のＩＮＤＥＸを対応させたＩＮＤＥＸテーブルと、図６１に示すＩＮＴＥＲ（フレーム間符号化）の非ＬＡＳＴ係数のＲＵＮ，ＬＥＶＥＬにリバーシブル符号の符号語（ＶＬＣ＿ＣＯＤＥ）のＩＮＤＥＸを対応させたＩＮＤＥＸにテーブルと、図６２に示すＩＮＴＲＡ，ＩＮＴＥＲ共通のＬＡＳＴ係数のＲＵＮ，ＬＥＶＥＬにリバーシブル符号の符号語（ＶＬＣ−ＣＯＤＥ）のＩＮＤＥＸを対応させたＩＮＤＥＸテーブルと、図６３，図６４に示すＩＮＤＥＸ値と符号語を対応させた符号語テーブルを持っている。
【０１７２】
下位階層可変長符号化器８０２の動作を、図５７のフローチャートに従って説明する。まず、モード情報に基づいて、ＩＮＴＥＲの時はＩＮＴＲＡの非ＬＡＳＴ係数とＬＡＳＴ係数の符号語テーブル，ＩＮＴＥＲの時はＩＮＴＥＲの非ＬＡＳＴ係数とＬＡＳＴ係数のＩＮＤＥＸテーブルを選択する。（Ｓ１０１）
次に、符号語テーブルを用いて符号化するにあたって、（ＲＵＮ，ＬＥＶＥＬ）がＩＮＤＥＸテーブルのＲＵＮの最大値Ｒ＿ｍａｘ，ＬＥＶＥＬの最大値Ｌ＿ｍａｘと比較して、ＩＮＤＥＸテーブル内に存在するかどうか確認する。（Ｓ１０２）
もし、存在した場合、図６０〜図６２のＩＮＤＥＸテーブルを引きＩＮＤＥＸの値が０であるかどうかチェックを行ない（Ｓ１０４），０でなければ、図１２の符号語テーブルのＩＮＤＥＸの符号語を出力する。符号語テーブルの符号語の最終ビットの“ｓ”は、ＬＥＶＥＬの符号を表し、“ｓ”が“０”のときは、ＬＥＶＥＬの符号は正、“ｓ”が“１”のときは、ＬＥＶＥＬの符号は負である（Ｓ１０５）。ＩＮＤＥＸの値が０であった場合と、範囲外であった場合は、符号語テーブルに存在しない係数で、図６７に示すようにＥＳＣＡＰＥ符号を続けて、ＬＡＳＴ係数かどうかの１ビットと図６５および図６６のＲＵＮとＬＥＶＥＬの絶対値を固定長符号化し、この固定長符号化部の先頭に２つのエスケープのうち“００００１”を付加し、末端にもエスケープ符号を付加する。図６３，図６４のＩＮＤＥＸ値が０の符号語が、エスケープ符号であり、エスケープ（ＥＳＣＡＰＥ）符号として用いられるＶＬＣ＿ＣＯＤＥの最終ビット“ｓ”はＬＥＶＥＬの符号を表し、“ｓ”が“０”ならばＬＥＶＥＬは正、“１”ならばＬＥＶＥＬは負である（Ｓ１０６）。
【０１７３】
以上説明したように、ＩＮＤＥＸテーブルを用いることにより、ＥＳＣＡＰＥ符号を用いた場合にも、探索することなく効率的に、符号語を出力できる。下位階層可変長復号化器８０２は、図６８，図６９の復号値テーブルを備えており、可変長符号を復号した結果得られるＩＮＤＥＸ値に基づいて動作する。
【０１７４】
次に、下位階層可変長復号化器の動作を、図５８と図５９のフローチャートに従って説明する。図５８は、順方向の場合の復号動作である。まず、可変長復号を順方向復号する（Ｓ２０１）。次に、復号して得られたＩＮＤＥＸ値が０かどうかをチェックし（Ｓ２０２）、０以外の場合は、上位階層のモード情報に基づいて、図６８の復号値テーブルのＩＮＴＲＡとＩＮＴＥＲを選択して、ＬＡＳＴ，ＲＵＮ，ＬＥＶＥＬの復号値を得る（Ｓ２０３）。ＩＮＤＥＸ値が０の場合は、ＥＳＡＰＥ符号であるから、次に続く固定長符号を復号して、ＬＡＳＴ，ＲＵＮ，ＬＥＶＥＬの復号値を得る（Ｓ２０４）。つづいて、末尾のＥＳＣＡＰＥ符号を復号する（Ｓ２０５）。ＬＥＶＥＬの正負は、符号語の最終ビットが“０”ならば、正であり、“１”ならば、負である（Ｓ２０６）。
【０１７５】
図５９は、逆方向の場合の復号動作である。まず、符号語の最初の１ビットで、ＬＥＶＥＬの正負を決定する。もし、“０”ならば、正であり、“１”ならば負である（Ｓ３０１）。次に可変長符号を逆方向復号する（Ｓ３０２）。逆方向復号して得られたＩＮＤＥＸ値が０かどうかチェックし（Ｓ３０３）、０以外の場合は、上位階層のモード情報に基づいて、図６８の復号値テーブルのＩＮＴＲＡとＩＮＴＥＲを選択して、ＬＡＳＴ，ＲＵＮ，ＬＥＶＥＬの復号値を得る（Ｓ３０４）。ＩＮＤＥＸ値が０の場合には、ＥＳＣＡＰＥ符号であるから、次に続く固定長符号を復号して、ＬＥＶＥＬ，ＲＵＮ，ＬＡＳＴの復号値を得る（Ｓ３０５）。つづいて、先頭のＥＳＣＡＰＥ符号を復号する（Ｓ３０６）。以上説明したように、復号値テーブルを用いることで、ＥＳＣＡＰＥ符号を使った場合にも記憶量を節約し、効率的に双方向に復号することができる。
【０１７６】
上記第１０実施形態におけるデコードプロセスの補強例について説明する。すなわち、動画像多重化部７０３での動画像符号化方式のシンタクスが図５３に示すシンタクスである場合の復号値判定方法について説明する。
【０１７７】
フレーム内符号化フレームでは、まず、上位階層のモード情報１とＩＮＴＲＡＤＣが復号される。もし、誤りが発見されて全てを復号できない場合は、誤りの生じた同期区間のマクロブロックの復号値を全てＮｏｔ Ｃｏｄｅｄとする。もし、符号化が最初のフレームで時間的に前のフレームが存在しない場合は、灰色あるいは特定の色でマクロブロックを埋める。
【０１７８】
下位階層では、モード情報２が誤りのため復号できない場合、下位階層の符号化データを全て放棄し、上位階層の復号値を書き換えて、Ｎｏｔ Ｃｏｄｅｄとするか、ＩＮＴＲＡ ＤＣのみで表示する。また、誤りのためＤＣＴ係数を放棄したマクロブロックは、Ｎｏｔ Ｃｏｄｅｄとするか、ＩＮＴＲＡ ＤＣのみで表示する。
【０１７９】
また、ＩＮＴＲＡモードのマクロブロックで、ＡＣ−ＤＣＴ係数の予測を行なっているマクロブロックがある場合は、可変長符号は復号できるが、周辺のマクロブロックからの予測を行なっているので、周辺のマクロブロックが復号できていないために復号できない場合、Ｎｏｔ Ｃｏｄｅｄとする。
【０１８０】
フレーム間符号化フレームでは、まず、上位階層のモード情報１と動きベクトルが復号される。もし、誤りが発見されて全てを復号できない場合は、誤りが生じた同期区間のマクロブロックの復号値は全てＮｏｔ Ｃｏｄｅｄとされる。もし、全て復号できた場合、同期符号ＭＭが存在することを確認し、もし、存在しなければ、やはり、同期区間のマクロブロックの復号値を全てＮｏｔ Ｃｏｄｅｄとする。
【０１８１】
下位階層では、モード情報２及びＩＮＴＲＡ ＤＣで誤りのため復号されない場合、下位階層の符号化データを放棄し、Ｎｏｔ Ｃｏｄｅｄと表示するか、または、前フレームからの動き補償のみでの表示（ＭＣ Ｎｏｔ Ｃｏｄｅｄ）に上位階層の復号値を書き換える。また、誤りのためＤＣＴ係数を放棄したマクロブロックは、ＭＣ Ｎｏｔ Ｃｏｄｅｄとする。
【０１８２】
また、ＩＮＴＲＡモードのマクロブロックで、ＡＣ−ＤＣＴ係数の予測を行なっているマクロブロックがある場合は、可変長符号は復号できるが、周辺のマクロブロックからの予測を行なっているので、周辺のマクロブロックが復号できていないために復号できない場合、Ｎｏｔ Ｃｏｄｅｄとする。
【０１８３】
図７０はフレーム間符号化フレームのＡＣ−ＤＣＴ係数部の詳しい復号値判定方法を示している。
【０１８４】
まず、図７０（ａ）に示すように順方向復号結果と逆方向復号結果で誤りが検出されるマクロブロックの位置（誤り検出位置）が交差しない場合は、誤りが検出されなかったマクロブロックの復号結果のみを復号値として使用し、二つの誤り検出位置の復号結果は復号値として使用せず、上位階層のモード情報の復号結果に基づいて、ＩＮＴＲＡマクロブロックについては、前フレームをそのまま表示するように、ＩＮＴＲＡマクロブロックについては、前フレームより動き補償のみで表示するように、上位階層の復号結果を書き換える。
【０１８５】
また、図７０（ｂ）に示すように順方向復号結果と逆方向復号結果の誤りの検出位置が交差する場合は、順方向復号の結果で誤りが検出されたマクロブロックまでの復号値は、順方向復号結果を使用し、それ以降の部分は、逆方向復号結果を復号値として使用する。
【０１８６】
なお、逆方向復号結果を優先して、逆方向復号の結果で誤りが検出されたマクロブロックまでの復号値は、逆方向復号結果を使用し、それ以前の部分は、順方向復号結果を復号値として使用してもよい。
【０１８７】
さらに、図７０（ｃ）に示すように同一のマクロブロックで順方向復号結果および逆方向復号結果の両方に誤りが検出される場合は、誤り検出位置のマクロブロックの復号値は放棄し、復号値として使用せずに、上位階層のモード情報の復号結果に基づいて、ＩＮＴＲＡマクロブロックについては、前フレームをそのまま表示するように、ＩＮＴＥＲマクロブロックについては、前フレームよりの動き補償のみで表示するように、上位階層の復号結果を書き換え、それ以後のマクロブロックに対する復号値は逆方向の復号結果を使用する。
【０１８８】
また、図７０（ｄ）に示すようにスタッフィング符号で誤りが検出され、逆方向に復号できない場合は、順方向復号のみ復号を行ない、誤りが検出された場合、誤り検出位置以降のマクロブロックは、上位階層のモード情報の復号結果に基づいて、ＩＮＴＲＡマクロブロックについては、前フレームをそのまま表示するように、ＩＮＴＥＲマクロブロックについては、前フレームよりの動き補償のみで表示するように、上位階層の復号結果を書き換え、それ以後のマクロブロックに対する復号値は逆方向の復号結果を使用する。
【０１８９】
図７０の復号値判定方法では、マクロブロックを単位に復号値の判定をおこなったが、ブロックの単位でも、符号語の単位で判定をおこなってもよい。
【０１９０】
以上実施形態で説明したように、本発明の可変長符号化／復号化装置は、従来と比較して、より少ない計算量と記憶量で十分であり、効率的である。
【０１９１】
具体的には、動画像符号化／復号化装置にも適用することが可能であり、より少ない計算量と記憶量で効率的な動画像符号化／復号化装置を提供することができる。
【０１９２】
【発明の効果】
以上、説明したように本発明によれば、より少ない計算量と記憶量で効率的に符号化／復号化することができる順方向にも逆方向にも復号可能な可変長符号化／復号化装置を提供することができる。
【０１９３】
また、本発明によれば無駄なビットパターンを減らして符号化効率を高め、しかもスタッフィング符号を用いて同期区間を一定区間を単位に設定する場合でも、順方向にも逆方向にも復号可能な可変長符号化／復号化装置を提供することができる。
【０１９４】
また、本発明における符号語の構成方法によれば、より幅広い情報シンボルの確率分布に対応でき、従来適用することのできなかったような情報シンボル数が大きな符号化に対しても適用が可能となる。具体的には、動画像符号化／復号化システムにも適用することが可能であり、誤りに強い動画像符号化／復号化システムを提供することができる。
【０１９５】
しかも、この符号語の構成法によれば符号化効率が向上し、また符号語のビットパターンの自由度が高い符号設計ができるため、同期符号を用いて同期区間の設定を行なう場合でも、同期符号と符号語のビットパターンが一致することによる疑似同期の問題を回避することができる。
【０１９６】
さらに、本発明によれば順方向および逆方向の復号結果から復号値を判定する際、順方向および逆方向復号を行なう際の誤り検出結果に応じて、符号化データの復号値を判定することにより、リバーシブル符号を伝送路誤りに対して効果的に復号することが可能となる。
【０１９７】
また、本発明によれば可変長符号の符号語の誤り検出機能を有する双方向復号化手段で復号処理を行なう際に、順方向復号処理での誤り検出結果に応じて符号化データの復号処理を切り替えることで、順方向と逆方向の復号処理を共用することができることから、回路規模や演算量を大幅に増加させることなく、伝送路誤りに対して効果的に復号を行なうことができる。
【０１９８】
さらに、本発明によれば符号化側で情報シンボルを重要度に応じて階層化した上で可変長符号化を行ない、同期区間毎に階層毎の符号化データを作成し、かつ階層毎の符号化データを多重化することにより、情報シンボルのシンタクスによらず順方向にも逆方向にも復号可能であって、より誤りに強い可変長符号化／復号化を行なうことが可能となる。
【図面の簡単な説明】
【図１】本発明の第１実施形態に係る可変長符号化／復号化装置の構成を示すブロック図。
【図２】図１における符号語テーブル作成部を示すブロック図。
【図３】図２における符号語構成部での第１の符号語構成方法を説明する図。
【図４】第１の符号構成方法で構成した符号語から作成した順方向および逆方向の復号木を示す図。
【図５】図２における符号語構成部での第２の符号語構成方法を説明する図。
【図６】図２における符号語構成部での第３の符号語構成方法を説明する図。
【図７】図２における同期区間設定部での同期区間設定方法を説明する図。
【図８】スタッフィング符号の第１の例を示す図。
【図９】スタッフィング符号の第２の例を示す図。
【図１０】図２における復号値判定部の動作を説明する図。
【図１１】本発明の第２実施形態に係る動画像符号化／復号化システムの概略構成を示すブロック図。
【図１２】第２実施形態に係る動画像符号化／復号化システムにおける符号化データのシンタックスを示す図。
【図１３】図１１における動画像多重化部および動画像多重化分離部の構成を示すブロック図。
【図１４】第２実施形態におけるＩＮＴＲＡおよびＩＮＴＥＲの非ＬＡＳＴ係数の符号語テーブルの一部を示す図。
【図１５】第２実施形態におけるＩＮＴＲＡおよびＩＮＴＥＲの非ＬＡＳＴ係数の符号語テーブルの他の一部を示す図。
【図１６】第２実施形態におけるＩＮＴＲＡおよびＩＮＴＥＲの非ＬＡＳＴ係数の符号語テーブルのさらに別の一部を示す図。
【図１７】第２実施形態におけるＩＮＴＲＡおよびＩＮＴＥＲのＬＡＳＴ係数の符号語テーブルの一部を示す図。
【図１８】第２実施形態におけるＩＮＴＲＡおよびＩＮＴＥＲのＬＡＳＴ係数の符号語テーブルの他の一部を示す図。
【図１９】第２実施形態におけるエスケープ符号の符号語テーブルを示す図。
【図２０】第２実施形態における符号語テーブルにない符号語の符号化方式を示す図。
【図２１】第２実施形態における最大の零ランの場合を示す図。
【図２２】第２実施形態における復号値判定部の動作を説明するためたの図。
【図２３】本発明の第３実施形態に係る可変長符号化／復号化装置の構成を示すブロック図。
【図２４】図２３における符号語構成部での第１の符号語構成方法を説明する図。
【図２５】図２３における符号語構成部での第２の符号語構成方法を説明する図。
【図２６】順方向符号語テーブルと逆方向符号語テーブルを共通化した双方向符号語テーブルの構成を示す図。
【図２７】復号化器に誤り検出部を付加した場合の構成を示すブロック図。
【図２８】第３実施形態における復号化方法を説明する図。
【図２９】第３実施形態における復号値判定部の復号値判定方法を説明するための図。
【図３０】本発明の第４の実施形態に係る可変長符号化／復号化装置の構成を示すブロック図。
【図３１】図３０における一つの階層の可変長符号化部の構成を示すブロック図。
【図３２】図３０における一つの階層の可変長復号化部の構成を示すブロック図。
【図３３】図３０におけるデータ階層化部および多重化部でのデータ階層化および多重化の例を示す図。
【図３４】図３２における復号値判定部の復号値判定方法の例を示す図。
【図３５】第４実施形態に係る可変長符号化／復号化装置を図１１の動画像符号化／復号化システムに組み込んだ場合の動画像多重化部および動画像多重化分離部での動画像符号化方式のシンタックスの第１の例を示す図。
【図３６】第４実施形態に係る可変長符号化／復号化装置を図１１の動画像符号化／復号化システムに組み込んだ場合の動画像多重化部および動画像多重化分離部での動画像符号化方式のシンタックスの第２の例を示す図。
【図３７】第４実施形態に係る可変長符号化／復号化装置を図１１の動画像符号化／復号化システムに組み込んだ場合の動画像多重化部および動画像多重化分離部での動画像符号化方式のシンタックスの第３の例を示す図。
【図３８】第４実施形態に係る可変長符号化／復号化装置を図１１の動画像符号化／復号化システムに組み込んだ場合の動画像多重化部および動画像多重化分離部での動画像符号化方式のシンタックスの第４の例を示す図。
【図３９】第４実施形態における動きベクトルの符号化テーブルを示す図。
【図４０】第４実施形態における動きベクトルの符号化テーブルを示す図。
【図４１】第４実施形態における動きベクトルの符号化テーブルを示す図。
【図４２】第４実施形態における動きベクトルの符号化テーブルを示す図。
【図４３】第４実施形態における動きベクトルの符号化テーブルを示す図。
【図４４】第４実施形態における動きベクトルの符号化テーブルを示す図。
【図４５】第４実施形態における動きベクトルの符号化テーブルを示す図。
【図４６】第４実施形態における動きベクトルの符号化テーブルを示す図。
【図４７】第４実施形態における動きベクトルの一次元予測を説明するための図。
【図４８】図３５のシンタクスの場合の復号値判定方法を説明する図。
【図４９】図３６のシンタクスの場合の復号値判定方法を説明する図。
【図５０】本発明の第６実施形態に係る符号化／復号化装置であり、図１における符号語テーブル作成部の詳細構成を示すブロック図。
【図５１】本発明の第７実施形態に係る動画像符号化／復号化システムの概略構成を示すブロック図。
【図５２】図５０における動画像多重化部および動画像多重化分離部の構成を示すブロック図。
【図５３】第７実施形態に係る動画像多重化部および動画像多重化分離部での動画像符号化方式のシンタクスの例を示す図。
【図５４】本発明の第８実施形態に係る情報源符号化器の一例を示すブロック図。
【図５５】本発明の第８実施形態に係る情報源復号化器の一例を示すブロック図。
【図５６】本発明の第９実施形態に係る可変長符号化／復号化装置が組み込まれるシステムの一例を示す図。
【図５７】本発明の第１０実施形態の可変長符号化器の動作を示すフローチャート。
【図５８】第１０実施形態の順方向可変長復号化器の動作を示すフローチャート。
【図５９】第１０実施形態の逆方向可変長復号化器の動作を示すフローチャート。
【図６０】ＩＮＴＲＡ係数のＩＮＤＥＸテーブルを示す図。
【図６１】ＩＮＴＥＲ係数のＩＮＤＥＸテーブルを示す図。
【図６２】ＬＡＳＲ係数のＩＮＤＥＸテーブルを示す図。
【図６３】第１０実施形態における符号語テーブルを示す図。
【図６４】第１０実施形態における符号語テーブルを示す図。
【図６５】第１０実施形態におけるＲＵＮの固定長符号語テーブルを示す図。
【図６６】第１０実施形態におけるＬＥＶＥＬの固定長符号語テーブルを示す図。
【図６７】符号語テーブルにない符号語の符号化方式を示す図
【図６８】第１０実施形態における復号値テーブル
【図６９】第１０実施形態における復号値テーブル
【図７０】復号値判定部の動作を説明するための図
【図７１】リバーシブル符号の一般的な復号方法を示す図。
【図７２】通常の可変長符号の説明図。
【図７３】従来のリバーシブル符号の説明図。
【図７４】従来のスタッフィング符号を示す図。
【図７５】従来の逆復号ができない情報シンボルのシンタックスを説明するための図。
【符号の説明】
５１ 双方向符号語テーブル
５２ アドレス変換器
５３ 識別信号
６１ 誤り検出部
１０１ 符号語テーブル作成部
１０２ 符号語テーブル
１０３ 符号化器
１０４ 同期区間設定部
１０５，２０５，７０５ 伝送系または蓄積系
１０６ 同期区間検出部
１０７ バッファ
１０８ 順方向復号化器
１０９ 逆方向復号化器
１１０ 復号値判定部
１１１ 順方向符号語テーブル
１１２ 逆方向符号語テーブル
１１３，２１３ 符号化部
１１４，１２１，２１４ 復号化部
１２２ 符号化データ切替器
１２３ 符号語テーブル切替器
１２４，６２ 復号化器
２０１ 符号化テーブル作成部
２０２ データ階層化部
２０３ 階層別可変長符号化部
２０４ 多重化部
２０６ 多重化分離部
２０７ 階層別復号化部
２０８ データ合成部
４０１ ブロック化回路
４０２ 減算器
４０３ モード選択スイッチ
４０４ ＤＣＴ回路
４０５ 量子化回路
４０６ 逆量子化回路
４０７ ＩＤＣＴ回路
４０８ 加算器
４０９ フレームメモリ
４１０ 動き補償回路
４１１ スイッチ
４１２ モード選択回路
５０１ 逆量子化回路
５０２ ＩＤＣＴ回路
５０３ 加算器
５０４ モード判定回路
５０５ モード切替スイッチ
５０６ フレームメモリ
７０１ 符号化制御部
７０２ 情報源符号化器
７０３ 動画像多重化器
７０４ 送信バッファ
７０６ 受信バッファ
７０７ 動画像多重化分離器
７０８ 情報源復号化器
７０９ 動画像符号化器
７１０ 動画像復号化器
９０１ 上位階層可変長符号化器
９０２ 下位階層可変長符号化器
９０３ 多重化器
９０４ 多重化分離器
９０５ 上位階層可変長復号化器
９０６ 下位階層可変長復号化器
１００１ パーソナルコンピュータ
１００２ ワークステーション
１００３，１００４ 無線機
１００５，１００６ カメラ[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a variable length encoding / decoding device used for compression encoding / decoding of moving image signals and the like, and in particular, a variable length encoding / decoding device capable of decoding from both forward and backward directions. The present invention also relates to a recording medium for recording data or programs used in this apparatus.
[0002]
[Prior art]
A variable-length code is based on the frequency of occurrence of a symbol, and assigns a short code length on average by assigning a short code length code to frequently appearing symbols and a long code length code to symbols that do not appear rarely. This is a code system designed to be Therefore, when a variable length code is used, the amount of data can be greatly compressed as compared with the data before encoding. As a method for constructing such a variable length code, an optimum Huffman algorithm for a non-recording information source is known.
[0003]
As a general problem of variable-length codes, when an error is mixed in a code due to an error in the transmission path of the encoded data or other reasons, the influence of the data after the error is mixed propagates to the decoding device. It will be impossible to decrypt correctly. In order to avoid this problem, when there is a possibility that an error occurs in the transmission path, a method is generally adopted in which a synchronization code is inserted at a certain interval to prevent error propagation. A bit pattern that does not appear in a combination of variable length codes is assigned to the synchronization code. According to this method, even if an error occurs in the encoded data and decoding becomes impossible, it is possible to prevent error propagation and perform correct decoding by finding the next synchronization code.
[0004]
However, even when a synchronization code is used, decoding is performed on encoded data between a location where an error occurs and decoding cannot be performed correctly as shown in FIG. 71 (a) until a location where the next synchronization code is found. I can't.
[0005]
Therefore, the variable length code is changed from the normal configuration shown in FIG. 72 as shown in FIG. 73, and not only the property that can be decoded from the normal forward direction as shown in FIG. A method for making a decodable code configuration is known. Further, since such a code can read the encoded data from the reverse direction, it can also be used for reverse reproduction in a storage medium such as a disk memory for storing the encoded data. A variable length code that can be decoded not only in the forward direction but also in the reverse direction is referred to as a reversible code here.
[0006]
An example of a reversible code is described in, for example, “Reversible Variable Length Coding Method” of Japanese Patent Laid-Open No. 5-300027. The reversible code of this known example is such that each codeword has a longer code length as shown in FIG. 73 at the end of the codeword of the Huffman code, which is a variable length code that can be decoded from the forward direction shown in FIG. This is a variable length code that can be decoded in the reverse direction by adding bits under conditions that do not match the end of other codewords. However, since this reversible code adds bits to the end of the codeword of a variable-length code that can be decoded only in the forward direction, the number of wasted bits increases and the average code length increases. As a result, the encoding efficiency is greatly reduced as compared with a variable-length code that can be decoded only in the forward direction.
[0007]
Further, the conventional reversible code has a problem that backward decoding cannot be performed if the synchronization interval is set in units of a certain interval. For example, ITU-T H.I. In H.263 (1996), the synchronization code is aligned every 8 bits (= 1 byte), that is, the position where the synchronization code can be inserted can be set to a position of 1 byte unit. In order to realize such synchronization interval alignment, a stuffing code as shown in FIG. 34 is inserted before the synchronization code. In such a case, the conventional reversible code has a problem that backward decoding cannot be performed due to the stuffing code.
[0008]
Furthermore, when simply decoding a reversible code, it requires twice the circuit scale and the amount of calculation as much as decoding in the reverse direction, compared to a variable length code decoding apparatus that can decode only in the normal forward direction. There was a problem of doing.
[0009]
Another problem with the conventional reversible code is that there is a case where decoding in the reverse direction is impossible due to the syntax of the input information. For example, in the case of an information symbol having the syntax of FIG. 75 (a), the code of symbol B is determined by symbol A. Since encoded data obtained by encoding an information symbol having such a syntax cannot decode B unless A can be decoded first as shown in FIG. 75B, it cannot be decoded from the reverse direction.
[0010]
As an encoding method applied when the number of information symbols is large, there is a method using an escape code. This method using an escape code has a code word table for a small number of information symbols with high appearance frequency as a code word table, and encodes the majority of information symbols with low appearance frequency with a combination of escape codes and fixed-length codes. It is. Also in the method using the escape code, a method has been proposed in which escape codes are added to the beginning and the end of a fixed-length code, as in the case of a variable-length code that can be decoded in both the forward and backward directions.
[0011]
In a conventional encoding / decoding method using an escape code, in order to distinguish between a codeword existing in the codeword table and a codeword encoded using the escape code, the codeword is stored in the codeword table by some method. There was a need to search for the existence.
[0012]
[Problems to be solved by the invention]
An object of the present invention is to provide a variable-length encoding / decoding device capable of efficient encoding / decoding with a small calculation amount and storage amount and capable of decoding in both forward and backward directions. .
[0013]
That is, as described above, the reversible code according to the prior art, that is, the variable length code that can be decoded from the forward direction or the reverse direction, adds a bit to the end of the code word of the variable length code that can be decoded only from the forward direction. As a result, a reversible code is formed, which increases the number of useless bit patterns, lengthens the average code length, and significantly reduces the coding efficiency compared to a variable-length code that can be decoded only in the forward direction. In addition, if the synchronization interval is set in units of a fixed interval using stuffing code, there is a problem that reverse decoding of the reversible code cannot be performed due to the stuffing code, and furthermore, when reversible code is simply decoded. Compared with a variable length code decoding device capable of decoding only in the normal forward direction, the circuit is doubled for decoding in the reverse direction. It requires diagrammatic and calculation amount, also depending on the syntax of information symbols input there is a problem that there are cases it is impossible to decode from the opposite direction.
[0014]
The present invention improves the coding efficiency by reducing useless bit patterns, and is a variable-length code that can be decoded in both the forward and reverse directions even when the synchronization interval is set in units of a fixed interval using a stuffing code. An object is to provide an encoding / decoding device.
[0015]
In addition, the present invention reduces useless bit patterns to increase coding efficiency, and can perform decoding efficiently by reducing the circuit scale and the amount of computation by sequentially performing decoding processing. It is another object of the present invention to provide a variable-length encoding / decoding device that can also decode.
[0016]
It is another object of the present invention to provide a variable-length encoding / decoding device that can decode both forward and backward directions regardless of the syntax of an input information symbol and is more resistant to errors.
[0017]
[Means for Solving the Problems]
The variable length coding apparatus according to the present invention assigns a codeword having a code length corresponding to the occurrence probability of each information symbol to a plurality of information symbols, and codes a codeword corresponding to the input information symbol. In a variable-length coding apparatus that outputs data, hierarchization means for hierarchizing input information symbols according to importance, and can decode both forward and backward, and codewords are predetermined. A codeword table storing a plurality of codewords including codewords configured so as to identify codeword breaks according to the weights, corresponding to the information symbols, and hierarchized from the codeword table by the layering means A code word selection means for selecting a code word corresponding to the information symbol, and a code word selected for each hierarchy using a code word selected by the code word selection means. Together to create the data, and having a multiplexing means for multiplexing the encoded data of each hierarchical layer.
[0018]
Here, the “codeword weight” corresponds to the Hamming distance with respect to the minimum value or the maximum value of the codeword. When the code word is a binary code, the minimum value of the code word is all “0” and the maximum value is all “1”, so the weight of the code word is a number of “1” or “0”. It corresponds to. The position where the weight of the code word becomes a predetermined value indicates a code break in the variable length code.
[0019]
In this variable length coding apparatus, a code length, that is, a variable length code composed of a code word having a predetermined code length, is configured by a value that does not depend on the code word order relationship, that is, a code word weight. Therefore, since the variable length code can be recognized from both the forward direction and the reverse direction, the variable length code is a reversible code that can be decoded from both directions. Also, unlike a conventional reversible code that adds a bit to the end of a variable-length code that can be decoded only in the forward direction, a reversible code is constructed from the beginning without adding extra bits. A variable length code with few bit patterns and high coding efficiency can be obtained.
[0020]
Further, in this variable length coding apparatus, even when setting a synchronization interval in units of a certain interval, it is possible to decode a reversible code in the reverse direction by using a stuffing code that can be decoded in the reverse direction. That is, when the reversible code is reversely decoded, it is necessary to know the end of the reversible code (the head when viewed in the reverse direction). However, if a stuffing code that can be decoded in the reverse direction is used, the end of the reversible code is known. Therefore, reverse decoding of the reversible code can be realized.
[0021]
In this variable-length coding apparatus, the input information symbols are hierarchized, then variable-length coding is performed using the same coding table as before, and further, a synchronization interval is set for each layer. Therefore, it is possible to decode both forward and backward directions regardless of the syntax of the information symbol, and it is possible to perform variable-length coding that is more resistant to errors.
[0022]
In the variable length coding apparatus having the above configuration, among the plurality of codewords stored in the codeword table, the codeword can be decoded both in the forward direction and in the reverse direction, and the codeword is determined by a predetermined weight of the codeword. Codewords configured to be delimited are preferably decoded in the forward and backward directions, for example at least one of the beginning and end of the first codeword that can be decoded in both forward and backward directions. It is constructed by adding a possible second codeword, or at least one immediately before and immediately after each bit of the first codeword that can be decoded both in the forward direction and in the reverse direction. A second codeword that can also be decoded in the direction is added.
[0023]
By configuring the codeword in this way, it is possible to deal with a wider probability distribution of information symbols, so that the coding efficiency is improved and the code design with a high degree of freedom of the bit pattern of the codeword can be performed. Even when the synchronization interval is set using, the problem of pseudo-synchronization due to the coincidence of the synchronization code and the bit pattern of the codeword can be avoided.
[0024]
Moreover, in the variable length coding apparatus having the above configuration, among the plurality of codewords stored in the codeword table, decoding is possible both in the forward direction and in the backward direction, and the codeword is encoded with a predetermined weight of the codeword. Codewords that are configured to identify word breaks are codewords in which a fixed-length codeword is inserted between each bit of a codeword that can be decoded in both the forward and backward directions. There may be.
[0025]
In addition, among a plurality of code words stored in the code word table, the code word can be decoded both in the forward direction and in the reverse direction, and the code word delimiter can be identified by a predetermined weight of the code word. The code word is a code word in which a code word that can be decoded in both the second forward direction and the reverse direction is inserted between each bit of the code word that can be decoded in the first forward direction and the reverse direction. There may be.
[0026]
By configuring the codeword in this way, it is possible to deal with a wider probability distribution of information symbols, so that the coding efficiency is improved and the code design with a high degree of freedom of the bit pattern of the codeword can be performed. Even when the synchronization interval is set using, the problem of pseudo-synchronization due to the coincidence of the synchronization code and the bit pattern of the codeword can be avoided.
[0027]
A first variable length decoding device according to the present invention is a variable length decoding device corresponding to the first variable length coding device, which can decode both in the forward direction and in the reverse direction, and has a predetermined synchronization. In a variable-length decoding device that decodes encoded data composed of a variable-length code composed of codewords including a codeword in which a stuffing code that can be decoded in the reverse direction is inserted for each interval, a synchronization interval of encoded data is determined Sync period detecting means for detecting, forward decoding means for decoding the encoded data of the sync section detected by the sync section detecting means from the forward direction, and encoded data of the sync section detected by the sync section detecting means And a backward decoding means for decoding from the backward direction.
[0028]
The variable length decoding device may include a decoded value determining unit that determines a decoded value from the decoding results of the forward decoding unit and the backward decoding unit.
[0029]
Here, the synchronization interval detecting means detects the number of bits of the encoded data, for example, by decoding the stuffing code from the reverse direction.
[0030]
In addition, at least one of the forward decoding means and the backward decoding means may detect a codeword that is detected as an error when a codeword that does not exist in the codeword table of the variable-length code appears in the encoded data. An error is detected if a codeword that does not exist in the codeword table detects an error when it appears in the encoded data, or if the number of bits of the decoded encoded data does not match the number of bits of the transmitted encoded data To do.
[0031]
Further, at least one of the forward decoding means and the backward decoding means detects an error in the decoding process when the decoded value obtained by decoding the encoded data is inappropriate.
[0032]
On the other hand, the decoded value determining means determines that the decoding result of the forward decoding means and the backward decoding means is correct when an error is detected in at least one of the forward decoding means and the backward decoding means, for example. The estimated decoding result is used, and a portion that cannot be decoded by any decoding result is discarded.
[0033]
In particular,
(A) When the error is detected by both the forward decoding means and the backward decoding means and their error detection positions do not intersect, only the decoding result in which no error is detected is used as a decoded value. ,
(B) When the error is detected by both the forward decoding means and the backward decoding means, the errors are detected, and the positions where the errors are detected intersect, the forward decoding is performed. The decoding value for the codeword up to the position immediately before the position where the error was found by the converting means uses the forward decoding result, and the decoding value for the codeword thereafter uses the backward decoding result,
(C) If the error is detected by only one of the forward decoding means and the backward decoding means, the decoded value for the codeword up to immediately before the position where the error is detected is the forward value. Using the decoding result, the decoding value for the codeword after that uses the decoding result in the reverse direction,
(D) When the error is detected for the same codeword by the forward decoding means and the backward decoding means, the decoded value for the codeword at the position where the error is detected is discarded, and the subsequent The decoding value for the code word uses the decoding result in the reverse direction.
[0034]
The decoded value determining means may be configured to abandon all portions that may contain errors when an error is detected in at least one of the forward decoding means and the backward decoding means.
[0035]
In this variable length decoding device, when determining the decoding value from the decoding results of the forward decoding means and the backward decoding means having the error detection function of the code word of the variable length code, the forward decoding means and the backward direction By determining the decoded value of the encoded data according to the error detection result in the decoding means, the reversible code output from the variable length encoding device as described above is effectively decoded against the transmission path error. It becomes possible to do.
[0036]
In addition, since a code that can be decoded in the reverse direction is used as the stuffing code, the reverse decoding of the reversible code can be performed even when the synchronization interval is set for each fixed interval using the stuffing code.
[0037]
A second variable length decoding device according to the present invention is a code comprising a variable length code that can be decoded both in the forward direction and in the reverse direction, and a stuffing code that can be decoded in the reverse direction in each predetermined synchronization interval. In a variable length decoding device for decoding encoded data, a synchronization interval detecting means for detecting a synchronization interval of encoded data, and a forward direction and a backward direction for decoding encoded data in the synchronization interval detected by the synchronization interval detecting means And bi-directional decoding means capable of decoding in both directions.
[0038]
Here, the bidirectional decoding means detects an error in the decoding process when the decoded value obtained by decoding the encoded data is inappropriate in at least one of the forward decoding process and the backward decoding process. .
[0039]
As a decoding method of the bidirectional decoding means, the following method can be used.
[0040]
(A) After performing the forward decoding process, the backward decoding process is performed from the end of the encoded data.
[0041]
(B) When an error is detected in the forward decoding process, the forward decoding process is stopped, and the backward decoding process is performed from the end of the encoded data.
[0042]
(C) When an error is detected in the forward decoding process, the forward decoding process is restarted from a position separated by an arbitrary distance from the error position.
[0043]
(D) When an error is detected in the forward decoding process, the backward decoding process is performed from a position that is an arbitrary distance away from the error position to the error position.
[0044]
(E) In the forward decoding process, the backward decoding process is performed only when an error is detected at the end of the encoded data.
[0045]
This second variable length decoding device may also have a decoded value determining means for determining a decoded value from the decoding results of the forward decoding means and the backward decoding means.
[0046]
In the second variable length decoding device, when the decoding process is performed by the bidirectional decoding means having the error detection function of the code word of the variable length code, the encoding is performed according to the error detection result in the forward decoding process. By switching the data decoding process, it is possible to share the decoding processing means in the forward direction and the reverse direction, so that the reversible output from the variable length coding apparatus as described above can be performed without significantly increasing the circuit scale and the amount of calculation. It is possible to effectively decode the code against a transmission path error.
[0047]
A third variable length decoding apparatus according to the present invention is a variable length decoding apparatus corresponding to the third variable length encoding apparatus, and includes a codeword including a codeword that can be decoded both in the forward direction and in the reverse direction. In a variable-length decoding device that decodes encoded data that is created by a codeword corresponding to hierarchical information symbols and multiplexed, for each predetermined synchronization interval, Separating means for separating the encoded data into hierarchies, a synchronization section detecting means for detecting a synchronization section of the encoded data separated by the separating means, and a synchronization section detected by the synchronization section detecting means Forward decoding means for decoding the encoded data from the forward direction, reverse decoding means for decoding the encoded data of the synchronization interval detected by the synchronization interval detection means from the reverse direction, these forward decoding means, and And having a synthesizing means for synthesizing the decoding results obtained for each hierarchy by the direction decoding means.
[0048]
This third variable length decoding device may also have a decoded value determining means for determining a decoded value from the decoding results of the forward decoding means and the backward decoding means.
[0049]
By using this third variable length decoding apparatus together with the third variable length encoding apparatus, it is possible to decode in the forward direction and the reverse direction regardless of the syntax of the information symbol, and is more resistant to errors. A variable-length encoding / decoding device can be realized.
[0050]
Further, in the recording medium on which variable length encoded data that can be input / output by the first variable encoding / decoding device according to the present invention is recorded, it can be decoded both in the forward direction and in the reverse direction. Coded data including a codeword configured so that a delimiter of a codeword can be recognized by a predetermined weight, and data obtained by inserting a stuffing code that can be decoded in the reverse direction for each predetermined synchronization interval into the coded data. It is characterized by being recorded and readable.
[0051]
Furthermore, in a recording medium recording transform coefficient data that can be generated by performing orthogonal transform in units of blocks in an image encoding / decoding device to which the variable length encoding / decoding device according to the present invention is applied,
The transform coefficient data includes orthogonal transform coefficients other than the last of the block after a plurality of codes that can be decoded in both the forward and reverse directions for each of a plurality of coding modes of the image encoding / decoding device. Of these, the data corresponding to the one with the highest appearance frequency and the plurality of encoding modes of the image encoding / decoding device are provided in common and can be decoded both in the forward direction and in the reverse direction. In the fixed orthogonal codes, the data corresponding to the most frequently occurring one of the last orthogonal transform coefficients of the block and the transform coefficient data having a low appearance frequency are described in a fixed-length code. It is readable by recording data with a codeword that can be decoded in the forward or backward direction at the end and data in which a stuffing code that can be decoded in the reverse direction is inserted into the encoded data for each predetermined synchronization section. That thing And it features.
[0052]
In addition, a recording medium on which variable length encoded data that can be input / output by the variable encoding / decoding device according to the present invention can be decoded both in the forward direction and in the reverse direction. Recorded encoded data including codewords configured to recognize codeword breaks by weights, and data in which stuffing codes that can be decoded in the reverse direction are inserted into the encoded data for each predetermined synchronization interval. It is characterized by doing.
[0053]
Furthermore, in the recording medium on which transform coefficient data that can be generated by performing orthogonal transform in units of blocks in the image encoding / decoding apparatus according to the present invention is recorded, the transform coefficient data is included in the image encoding / decoding apparatus. A plurality of codewords that can be decoded in both the forward and backward directions for each of a plurality of encoding modes, respectively, corresponding to a high-frequency occurrence of orthogonal transform coefficients other than the last block A plurality of codewords that are provided in common for a plurality of encoding modes of the image encoding / decoding device and can be decoded in both forward and backward directions, among the last orthogonal transform coefficients of the block Describe the data corresponding to those with high appearance frequency and transform coefficient data with low appearance frequency with fixed-length codes, and add codewords that can be decoded in the forward and backward directions at the beginning and end Day When, it is characterized in that for recording, and data Suffingu inserting the code decodable in the opposite direction for each predetermined synchronization interval on the coded data.
[0054]
Also, a variable-length coding apparatus that assigns a code word having a code length corresponding to the occurrence probability of the information symbol to a plurality of information symbols, and outputs a code word corresponding to the input information symbol as coded data The recording medium on which the program used in the above is recorded is capable of decoding both in the forward direction and in the reverse direction, and includes a plurality of codewords configured so that the codeword breaks can be recognized by the predetermined weight of the codeword A code word table in which each code word is associated with an information symbol, a procedure for selecting a code word corresponding to the input information symbol from the code word table, and a procedure for selecting the code word Creating encoded data for each predetermined synchronization interval using the selected codeword and inserting a stuffing code that can be decoded in the reverse direction Ri is characterized by recording at least including programs, and procedures for setting the synchronization period to the codeword.
[0055]
Furthermore, the encoded data is composed of a variable-length code composed of codewords including codewords that can be decoded both in the forward direction and in the reverse direction, and a stuffing code that can be decoded in the reverse direction is inserted for each predetermined synchronization interval. A recording medium on which a program used in a variable-length decoding device that decodes is recorded is a procedure for detecting a synchronization interval of the encoded data, and decodes encoded data of the synchronization interval detected by the procedure from the forward direction. A program including at least a procedure and a procedure for decoding encoded data of the synchronization section detected by the procedure of detecting the synchronization section from the reverse direction is recorded.
[0056]
DETAILED DESCRIPTION OF THE INVENTION
Hereinafter, embodiments of the present invention will be described with reference to the drawings.
[0057]
(First embodiment)
FIG. 1 is a block diagram showing a configuration of a variable length encoding / decoding device according to a first embodiment of the present invention.
[0058]
The variable-length encoding / decoding apparatus according to the first embodiment is roughly composed of a codeword table creation unit 101, an encoding unit 113, a transmission system or storage system 105, and a decoding unit 114. First, the functions of these units will be briefly described. The codeword table creation unit 101 creates a codeword table based on the occurrence probability of information symbols, and the codeword table 102 in the coding unit 113 and the decoding unit 114. Codewords are sent to the forward codeword table 111 and the backward codeword table 112.
[0059]
Encoding section 113 encodes the information symbol into a variable length code, and outputs the variable length code as encoded data to transmission system or storage system 105. Decoding section 114 decodes the encoded data input via transmission system or storage system 105 to reproduce the original information symbol.
[0060]
Next, the detailed configuration and operation of each part of the first embodiment will be described.
[0061]
In the encoding unit 113, the input information symbol is input to the encoder 103. The encoding unit 113 includes a codeword table 102, an encoder 103, and a synchronization interval setting unit 104. The codeword table 102 stores information symbols created in advance by the codeword table creation unit 101 in correspondence with codewords of variable length codes. The encoder 103 selects and outputs a code word corresponding to the input information symbol from the code words stored in the code word table 102. The synchronization interval setting unit 104 collects the codewords selected by the encoder 103 for each synchronization interval, and further inserts a stuffing code that can be decoded in both the forward and backward directions, and generates encoded data for each synchronization interval. Output. This encoded data is sent to the decoding unit 114 through the transmission system or the storage system 105.
[0062]
The decoding unit 114 includes a synchronization interval detection unit 106, a buffer 107, a forward decoder 108, a backward decoder 109, a decoded value determination unit 110, a forward decoding table 111, and a backward decoding table 112. Composed. In the decoding unit 114, the synchronization interval of the encoded data is detected by the synchronization interval detection unit 106 from the encoded data input from the transmission system or the storage system 105, and the encoded data is stored in the buffer 107. The forward decoder 108 starts decoding from the beginning of the encoded data stored in the buffer 107. The backward decoder 109 starts decoding from the end of the encoded data stored in the buffer 107.
[0063]
The forward decoder 108 detects an error when a bit pattern that does not exist in the forward codeword table 111 appears in the encoded data, or when encoded data having a bit number different from the bit number of the buffer 107 is decoded. Is determined. Similarly, in the reverse decoder 109, when a bit pattern that does not exist in the reverse codeword table 112 appears in the encoded data, and when encoded data having a bit number different from the bit number of the buffer 107 is decoded, Judged as error detection.
[0064]
The decoded value determination unit 110 obtains a decoded value from the decoding result (referred to as the forward decoding result) obtained by the forward decoder 108 and the decoding result (referred to as the backward decoder) obtained from the backward decoder 109. And the final decoding result is output. FIG. 2 is a block diagram showing the configuration of the codeword table creation unit 101. The code selection unit 21 receives information on the occurrence probability of an information symbol, selects a code system having the smallest average code length from selectable code systems, and sends the selection result to the codeword configuration unit 22. The code word configuration unit 22 configures the code word of the code selected by the code word selection unit 21.
[0065]
The codeword configured by the codeword configuration unit 22 is a variable-length code (hereinafter, reversible code) that can be decoded both in the forward direction and in the reverse direction so that the code delimiter can be identified by a predetermined weight of the codeword. For example, these are the codes described in Japanese Patent Application No. 7-87772 and Japanese Patent Application No. 7-260383.
[0066]
In the first embodiment, the following method is used as a codeword construction method in the codeword construction unit 22 in order to deal with a wider probability distribution and increase the degree of freedom of the bit pattern of the codeword.
[0067]
FIG. 3 shows a first configuration method of the codeword of the reversible code in the codeword configuration unit 22. First, as shown on the left side of FIG. 3, two binary sequences (in this case, the weight is 0) in which the weight of each sequence (in this case, the number of “1”) is constant and has different weights in ascending order of the code length. And 1). Next, as shown in the center of FIG. 3, after adding “1” to the beginning and the end of each binary sequence to invert the bits of the binary sequence, these two binary sequences are converted into those shown in FIG. Synthesize as shown on the right.
[0068]
In this variable length code, the code length can be determined by counting the number of symbols at the beginning of each code. In the example of FIG. 3, if the first is “0”, the code delimiter (code length) is known when two “0” appear, and if the first is “1”, the code It has a code configuration that allows the delimiters to be identified. Further, in the variable length code shown in FIG. 3, the codewords corresponding to all the information symbols A to J are also decoded in the reverse direction of FIG. 4B on the leaves of the forward decoding tree shown in FIG. Since it is also assigned to the leaves of the tree, it can be understood that decoding is possible in both the forward and backward directions.
[0069]
FIG. 5 shows a second configuration method of the codeword of the reversible code in the codeword configuration unit 22. First, as shown on the left side of FIG. 5, a first reversible code and a second reversible code are prepared. Next, as shown in the center of FIG. 5, the first one codeword of the second reversible code is added to the end of all the codewords of the first reversible code. Similarly, after adding all the codewords of the second reversible code one by one to the end of all the codewords of the first reversible code, rearrangement is performed as shown on the right side of FIG. A new reversible code is constructed. By such a configuration method, the number A × B obtained by multiplying the number A of codewords of the first reversible code (A = 9 in this example) by the number of codewords B of the second reversible code (B = 3 in this example). (27 in this example) new reversible codes can be constructed.
[0070]
In decoding the reversible code by this configuration method, when decoding from the forward direction, the first reversible code is first decoded, and then the second reversible code is decoded. When decoding from the reverse direction, first, the second reversible code is decoded, and then the first reversible code is decoded. Therefore, it can be understood that decoding is possible both in the forward direction and in the reverse direction.
[0071]
In this example, the second reversible code is added at the end of the codeword of the first reversible code. However, the second reversible code may be added at the beginning of the first reversible code. A fixed-length code may be added to both the end and the beginning of the reversible code. In this example, the first reversible code and the second reversible code are different, but the same may be used. Furthermore, in the present embodiment, variable-length codes are used for both the first reversible code and the second reversible code, but either of them may be changed to a fixed-length code.
[0072]
FIG. 6 shows a third code configuration method of the reversible code in the codeword configuration unit 21. First, as shown on the left side of FIG. 6, a variable-length reversible code and a fixed-length reversible code are prepared. Next, as shown on the right side of FIG. 6, a fixed-length reversible code is added immediately after each bit of the code word of the reversible code. With this configuration method, when a K-bit fixed-length reversible code is used, the number of codewords can be increased by 2KH by setting the H-bit codeword to (K + 1) H bits. In this example, a fixed-length code is added immediately after each bit of a reversible codeword. However, a fixed-length code may be added immediately before each bit, and a fixed-length code may be added immediately before and after each bit. May be added.
[0073]
Next, a specific example of a method for setting a synchronization interval of encoded data in the synchronization interval setting unit 104 will be described with reference to FIG. FIG. 7A shows a method for setting a synchronization interval by inserting a synchronization code. In this case, the synchronization interval detecting unit 106 can detect the synchronization interval by detecting the first and last synchronization codes of the encoded data. Here, in order to set the insertable position of the synchronization code in a certain section unit (in this case, M bit unit), a 1 to M bit stuffing code is inserted at the end of the encoded data.
[0074]
FIG. 7B shows a method of describing the code amount of the synchronization section at the head of the encoded data, in other words, a method of inserting a pointer indicating the end position of the synchronization section. In this case, the synchronization interval detection unit 106 can detect the synchronization interval by first decoding the description of the code amount of the first synchronization interval of the encoded data. Since the description unit of the code amount is an M-bit unit, a 1-M bit stuffing code is inserted at the end of the encoded data.
[0075]
Here, the stuffing code used in FIGS. 7A and 7B is characterized in that it can be decoded at least in the reverse direction. FIG. 8 is a diagram showing a specific example of such a stuffing code, and this stuffing code is configured to be decodable both in the forward direction and in the reverse direction. In the case of this stuffing code, decoding is possible in both the forward and backward directions, so that when 1 bit appears, if “1” appears, it is delimited, otherwise “0” appears twice. It is a variable length code that can be decoded in the reverse direction from the end of the synchronization interval, and at the same time, the number of bits of the encoded data can be calculated. The calculation result of the number of bits of the encoded data is used for error detection described later.
[0076]
In this example, the stuffing code is inserted at the end of the encoded data. However, the stuffing code may be at the beginning of the encoded data or inside the encoded data. Good. The stuffing code may be a code that can be decoded only in the reverse direction as shown in FIG. In the case of this code, the code is such that it is delimited when “0” appears after decoding from the reverse direction.
[0077]
Next, a decoded value determination method in the decoded value determination unit 110 will be described with reference to FIG. As shown in FIG. 10 (a), when the position of the decoded word (error detection position) where an error is detected in the forward decoding result and the backward decoding result does not intersect, only the decoding result in which no error is detected is obtained. Used as a decoded value, the decoding results at the two error detection positions are not used as decoded values but are discarded. Also, as shown in FIG. 10B, when the error detection position of the forward decoding result and the backward decoding result intersect, the decoding value up to the error detection position of the forward decoding result uses the forward decoding result. The subsequent part uses the backward decoding result as a decoded value.
[0078]
Note that the backward decoding result is given priority, the backward decoding result is used as the decoded value up to the error detection position of the backward decoding result, and the forward decoding result is used as the decoded value for the previous part. Good.
[0079]
Also, as shown in FIG. 10C, when an error is detected only in the one-way decoding result of the forward decoding result and the backward decoding result (in this example, only the forward decoding result is detected). The forward decoding result is used as the decoding value up to the error detection position, and the backward decoding result is used as the decoding value for the subsequent portions. Furthermore, as shown in FIG. 10 (d), when an error is detected in both the forward decoding result and the backward decoding result for the same codeword, the decoded value for the codeword at the error detection position is discarded, and The backward decoding result is used as a decoded value for the subsequent codeword. It should be noted that when a bit pattern that does not exist as a code word is generated in the error determination in the forward decoder 108 and the backward decoder 109, the detection position is the place, and no error is detected by the above determination method. In some cases, if the number of bits decoded does not match the number of bits of encoded data in the synchronization interval, the first position of decoding is set as the error detection position.
[0080]
In the first embodiment, one example of the decoded value determination method based on the patterns of the four error detection positions has been shown. However, the decoded value determination unit performs an error in one or both of the forward decoding result and the backward decoding result. If it is detected, the forward decoding result or the backward decoding result estimated to be correct is used, and any part that cannot be decoded by any decoding result is abandoned. Nor.
[0081]
(Second Embodiment)
Next, an application example of the variable-length encoding / decoding device according to the present invention will be described as a second embodiment of the present invention. FIG. 11 is a conceptual diagram of a moving image encoding / decoding system in which a variable length encoding / decoding device according to a second embodiment of the present invention is incorporated. First, in the moving image encoder 709, the data encoded by the information source encoder 702 is subjected to variable length encoding and multiplexing by the moving image multiplexer 703 and smoothed by the transmission buffer 704. After being converted, the data is sent to the transmission system or storage system 705 as encoded data. The encoding control unit 701 controls the information source encoder 702 and the moving picture multiplexer 703 in consideration of the buffer amount of the transmission buffer 704.
[0082]
On the other hand, in the video decoder 710, the encoded data from the transmission system or the storage system 705 is stored in the reception buffer 706, and the video data demultiplexer 707 demultiplexes the encoded data and performs variable length decoding. Is sent to the information source decoder 708 to finally decode the moving picture information. Here, the variable length coding / decoding apparatus described in the first embodiment is applied to the moving picture multiplexer 703 and the moving picture demultiplexing separator 707.
[0083]
FIG. 12 shows the syntax of the moving image encoding method in the moving image multiplexer 703 and the moving image demultiplexer / separator 707 in the second embodiment. Reversible code in the lower layer, with the macroblock mode information, motion vector information and INTRA DC (DC component of DCT coefficients in intra-frame coding) information as the upper layer, DCT coefficient information other than INTRA DC as the lower layer Has been applied.
[0084]
FIGS. 13A and 13B are block diagrams showing configurations of the moving image multiplexer 703 and the moving image demultiplexer / separator 707 in the second embodiment.
[0085]
In the moving picture multiplexer 703 shown in FIG. 13A, information such as macroblock mode information, motion vector information, and INTRA DC among the data encoded by the information source encoder 702 in FIG. Is an upper layer, normal variable length coding is performed by the upper layer variable length encoder 901, and sent to the multiplexer 903. Of the data encoded by the information source encoder 702, DCT coefficients other than INTRA DC are encoded by the reversible code in the lower layer variable length encoder 903 and sent to the multiplexer 903. The multiplexer 903 multiplexes the upper layer encoded data and the lower layer encoded data, and sends the multiplexed data to the transmission buffer 704.
[0086]
On the other hand, in the moving picture decoder 707 shown in FIG. 13 (b), the upper layer encoded data and the lower layer encoded data are separated by the demultiplexing unit 904, and these are separated into the upper layer variable length decoder. Variable length decoding is performed by 905 and lower layer variable length decoder 906, respectively.
[0087]
FIGS. 14 to 16, FIGS. 17 to 18, and 19 are examples of codeword tables used in the lower layer variable length encoder 902. The code stored in this codeword table is a code in which a 2-bit fixed-length code is added to the end of the code created by the first codeword construction method of the reversible code in the codeword construction unit 22 described above.
[0088]
The information source encoder 702 scans each block of the quantized 8 × 8 DCT coefficients for each block, and performs LAST (0: non-zero coefficient not at the end of block, 1: last non-zero coefficient at the block) ), RUN (number of zero runs up to non-zero coefficient) and LEVEL (quantized value of coefficient) are obtained and sent to the moving picture multiplexing unit 703.
[0089]
The lower layer variable length encoder 902 in the moving picture multiplexing unit 703 is reversible to non-last coefficients, RUN, and LEVEL of INTRA (intraframe coding) and INTER (interframe coding) shown in FIGS. Code word table of non-LAST coefficients corresponding to code word (VLC_CODE) of code and reversible code code word (VLC_CODE) to INTRA and INTER LAST coefficients, RUN, and LEVEL shown in FIGS. It has a codeword table of the last coefficients that have been made.
[0090]
Based on the mode information, encoding is performed by selecting a codeword table of INTRA non-LAST coefficients and LAST coefficients for INTRA, and a codeword table of INTER non-LAST coefficients and LAST coefficients for INTER. “S” in the last bit of the code word represents the sign of LEVEL. When “s” is “0”, the sign of LEVEL is positive, and when “s” is “1”, the sign of LEVEL is negative. .
[0091]
As shown in FIG. 20, the coefficient that does not exist in this codeword table is a fixed length encoding of 1 bit of whether or not the LAST coefficient and the absolute value of RUN and LEVEL, and two escape codes at the head of this fixed length code part. Among them, “00001” is added, and an escape code is also added to the end. FIG. 19 is a codeword table of escape codes, where “s” of the last bit of VLC_CODE used as an escape (ESCAPE) code represents the code of LEVEL, and when “s” is “0”, LEVEL is positive, If “1”, LEVEL is negative.
[0092]
FIG. 21 shows a case where the zero run is maximized when this codeword table is used. Normally, a bit pattern that adds “1” after a zero run is selected as the synchronization code is selected. For example, ITU-TH. In H.263, a bit pattern in which “1” is added after 16 “0” is a synchronization code. In variable length coding of DCT coefficients, the zero run is maximized when the coefficient using the escape code is LEVEL is +64, and the next code using the escape code appears again, and 15 “0” s are consecutive. It is the greatest to do. Therefore, if a bit pattern having 16 consecutive “0” s is used as a synchronization code, there is no possibility that a pseudo synchronization code is generated.
[0093]
In the decoded value determination unit 110, a decoded value is obtained from the decoding result (referred to as the forward decoding result) obtained by the forward decoder 108 and the decoding result (referred to as the backward decoding result) obtained from the backward decoder 109. And the final decoding result is output.
[0094]
In the error determination in the forward decoder 108 and the backward decoder 109, when a bit pattern that does not exist as a code word is generated, and when an error is detected by a check bit or the like, the detection position is set as the location, If no error is detected by this determination method and the number of decoded bits does not match the number of bits of encoded data in the synchronization interval, the first position of decoding is set as the error detection position.
[0095]
FIG. 22 shows a decoded value determination method for the lower layer. First, as shown in FIG. 22A, no error is detected when the position (error detection position) of the macroblock (MB) where an error is detected does not intersect between the forward decoding result and the backward decoding result. Only the decoding result of the macroblock is used as a decoding value, the decoding result of the two error detection positions is not used as a decoding value, and the INTRA macroblock (intraframe coding is used based on the decoding result of the higher layer mode information. The decoding result of the higher layer is rewritten so that the previous frame is displayed as it is for the macroblock), and the motion compensation is displayed only for the INTER macroblock (macroblock that has been interframe coded) from the previous frame. .
[0096]
Also, as shown in FIG. 22B, when the error detection positions of the forward decoding result and the backward decoding result intersect, the decoded values up to the macroblock in which the error is detected in the forward decoding result are The forward decoding result is used, and the subsequent decoding result is used as the decoded value for the subsequent portions. In addition, the backward decoding result is prioritized, and the decoding value up to the macroblock in which an error is detected in the backward decoding result uses the backward decoding result, and the previous part decodes the forward decoding result. It may be used as a value.
[0097]
Further, as shown in FIG. 22C, when an error is detected only in the one-way decoding result of the forward decoding result and the backward decoding result (in this example, only the forward decoding result is detected). The decoding value for the macroblock after the error detection position uses the decoding result in the reverse direction. Furthermore, as shown in FIG. 22 (d), when an error is detected in both the forward decoding result and the backward decoding result in the same macroblock, the decoded value of the macroblock at the error detection position is discarded, Instead of using it as a decoded value, based on the decoding result of the higher layer mode information, the previous frame is displayed as it is for the INTRA macroblock, and the motion compensation from the previous frame is displayed only for the INTER macroblock. Furthermore, the decoding result of the upper layer is rewritten, and the decoding value for the subsequent macroblock uses the decoding result in the reverse direction.
[0098]
In the decoded value determination method of FIG. 22, the decoded value is determined in units of macroblocks, but may be determined in units of blocks or codewords. In the second embodiment, the present invention is applied to variable length coding of DCT coefficients. However, it goes without saying that other variable information symbols can be applied to variable length coding.
[0099]
(Third embodiment)
Next, a variable-length encoding / decoding device according to the third embodiment of the present invention will be described with reference to FIG.
[0100]
The variable length coding / decoding apparatus according to the third embodiment is roughly composed of a codeword table creation unit 101, a coding unit 113, a transmission system or storage system 105, and a decoding unit 121. First, the functions of these units will be briefly described. The codeword table creation unit 101 creates a codeword table based on the occurrence probability of information symbols, and the codeword table 102 in the coding unit 113 and the decoding unit 121. Codewords are sent to the forward codeword table 111 and the backward codeword table 112. Encoding section 113 encodes the information symbol into a variable length code, and outputs the variable length code as encoded data to transmission system or storage system 104. Decoding section 121 decodes encoded data input via transmission system or storage system 105 to reproduce the original information symbol.
[0101]
Next, the detailed configuration and operation of each part of the third embodiment will be described. The encoding unit 113 is the same as that of the first embodiment shown in FIG. That is, in the encoding unit 113, the input information symbol is input to the encoder 103. The codeword table 102 stores information symbols created in advance by the codeword table creation unit 101 in correspondence with codewords of variable length codes. The encoder 103 selects a code word corresponding to the input information symbol from the code words stored in the code word table 102. The synchronization interval setting unit 104 collects the codewords selected by the encoder 103 for each synchronization interval, and further inserts a stuffing code that can be decoded in both the forward and backward directions, and generates encoded data for each synchronization interval. Output. This encoded data is sent to the decoding unit 121 through the transmission system or the storage system 104.
[0102]
The decoding unit 121 includes a synchronization interval detection unit 106, a buffer 107, an encoded data switch 122, a decoder 124, a codeword table switch 125, a decoded value determination unit 110, a forward decoding table 111, and a backward decoding. The configuration table 112 is configured. The encoded data switch 122 and the codeword table switch 125 are switched under the control of the decoder 124.
[0103]
The decoding unit 121 inputs the encoded data to the decoder 124 while the synchronization interval detecting unit 106 checks the synchronization interval of the encoded data from the encoded data input from the transmission system or the storage system 105. At that time, the encoded data is also accumulated in the buffer 107. The decoder 124 starts decoding the input encoded data (referred to as forward decoding). Here, the decoder 124 detects an error when a bit pattern that does not exist in the forward codeword table 111 appears in the encoded data, or when encoded data having a bit number different from the bit number of the buffer 107 is decoded. Is determined.
[0104]
When it is determined that an error has been detected, the decoder 124 switches the codeword table from the forward codeword table 111 to the backward codeword table 112 by the codeword table switch 123 and simultaneously decodes by the encoded data switch 122. The input to the generator 124 is switched to the buffer 107 side. When the synchronization period detection unit 106 detects the next synchronization, the buffer 107 reads the accumulated encoded data from the end and outputs it to the decoder 124. The decoder 124 starts decoding the input encoded data (referred to as backward decoding). Here, similarly to the forward decoding, the decoder 124 encodes encoded data having a bit number different from the bit number of the buffer 107 when a bit pattern that does not exist in the backward codeword table 112 appears in the encoded data. Is decoded, it is determined that an error has been detected. The decoded value determination unit 110 determines a decoded value from a decoding result (referred to as a forward decoding result) obtained by forward decoding and a decoding result (referred to as a backward decoding result) obtained from backward decoding, The final decoding result is output.
[0105]
The code word table creation unit 101 in the third embodiment is basically the same as that in the first embodiment. As shown in FIG. 2, a code that can be selected by inputting information on the occurrence probability of an information symbol. A code selection unit 21 that selects the smallest average code length from the system and a codeword configuration unit 22 that configures a codeword of the code selected by the codeword selection unit 21 are included. However, in the present embodiment, the following method is used as a codeword construction method in the codeword construction unit 22 in order to deal with a wider probability distribution and increase the degree of freedom of the codeword bit pattern.
[0106]
FIG. 24 shows a first configuration method of a codeword of a reversible code in the codeword configuration unit 22. First, as shown on the left side of FIG. 24, a first reversible code is prepared. Next, as shown in the center of FIG. 24, for each codeword of the first reversible code, a (code length-1) -bit fixed length code surrounded by a broken line is prepared, and the right side of FIG. As shown in FIG. 5, a fixed-length code word is inserted bit by bit so as to be surrounded by a broken line between each bit of the first reversible code word. It can be seen that the variable-length code in FIG. 24 can be decoded in both the forward and reverse directions by decoding the fixed-length code for each bit while decoding the first reversible code. In the example of FIG. 24, one bit of a fixed-length code word is inserted between each bit of the code word of the first reversible code, but for each code word of the first reversible code, (code A fixed-length code of length-1) × n bits may be prepared, and a fixed-length code word may be added n bits between bits of the first reversible code.
[0107]
FIG. 25 illustrates a second configuration method of the codeword of the reversible code in the codeword configuration unit 22. As shown on the left side of FIG. 25, a first reversible code is prepared. Next, a second reversible code is prepared as shown surrounded by a broken line on the upper side of FIG. 25, and a broken line is inserted between each bit of the code word of the first reversible code as shown on the right side of FIG. A second reversible code is inserted as enclosed. It can be seen that the variable-length code in FIG. 25 can be decoded in both the forward and reverse directions by decoding the second reversible code for each bit while decoding the first reversible code. In the example of FIG. 25, the first reversible code and the second reversible code are shown as the same code, but different codes may be used.
[0108]
In the third embodiment, as shown in FIG. 26, a bidirectional codeword table 51 in which the forward codeword table 111 and the backward codeword table 112 in FIG. 23 are shared may be used. In this case, an identification signal 53 indicating the distinction between forward decoding and backward decoding is input from the decoder 124 in FIG. 23, and the address converter 52 is operated by this identification signal 53 to generate a bidirectional codeword. An address for reading the code word from the table 51 is generated. Then, the code value corresponding to this address is read from the bidirectional codeword table 51.
[0109]
In the third embodiment, an error detection function is added to the decoder 124 as shown in FIG. The error detection unit 62 connected to the decoder 61 monitors the internal state of the decoder 61, and outputs an error detection signal when this becomes abnormal. An abnormal internal state refers to the following state, for example.
[0110]
(1) When encoded data that does not exist in the codeword table is received.
[0111]
(2) When the length of the encoded data does not match the length of the data actually decoded by the decoder 61.
[0112]
(3) A decoded value can be obtained because a value exists in the codeword table, but the value is inappropriate. Specific examples of (3) include (3-1) when the decoded value exceeds the existing range, (3-2) when the number of data exceeds the upper limit, and (3-3) decoded before For example, a decoded value that does not match the decoded value is output.
[0113]
(4) When an error is detected by an error detection code or the like.
[0114]
Next, the decoding method in this 3rd Embodiment is demonstrated using FIG. In the above description, the forward and backward decoding processes in the decoder 124 are performed simultaneously in both the forward and backward directions as shown in FIG. 28 (a), and each detects an error. Until then, decryption was executed to obtain a decrypted value. The present embodiment is not limited to this, and decoding may be performed by a method as shown in FIGS. In FIGS. 28 (a), (b), (c), and (d), X represents an error detection point, an arrow pointing to the right side represents a forward decoding process, and an arrow shown on the left side represents a backward decoding process. Show.
[0115]
In the method shown in FIG. 28 (b), when a forward decoding process is performed and an error is detected, the forward decoding process is stopped and an error is detected in the backward direction from the end of the encoded data in the synchronization interval. Decoding processing is performed until is detected. In the method shown in FIG. 28C, when an error is detected in the forward decoding process, the forward decoding process is restarted from n bits ahead of the error detection position. This process is repeated until all the encoded data in the current synchronization interval is processed. At the end of the decoding process in the forward direction, if the number of encoded data in the synchronization interval does not match the number of data decoded by the decoding process, it is regarded as error detection, and the decoding process in the reverse direction from the end of the encoded data is performed. Repeat until an error is found. In the method shown in FIG. 28D, when an error is detected in the forward decoding process, the decoding process is performed once for n bits in the reverse direction from the n-bit ahead of the error detection position. Thereafter, the forward decoding process is performed from n bits ahead of the error detection position (the position where the backward decoding process is started). This process is repeated until all the encoded data in the current synchronization interval is processed. Finally, if the number of encoded data in the synchronization interval and the number of data decoded by the decoding process do not match, it is regarded as error detection, and the reverse decoding process is performed from the end of the encoded data until an error is found. . Next, the determination method of the decoded value in the decoded value determination part 111 in this 3rd Embodiment is demonstrated using FIG. It should be noted that the following method can also be used to select a decoded value to be finally used by the decoded value determination unit 111. The decoded value determination method shown in FIG. 29A is a method in which when an error is detected, decoding is not performed thereafter. That is, when an error is detected, subsequent decoding may be erroneously decoded. Therefore, a decoded value that may cause such erroneous decoding is not used, and only the decoded value up to the error detection position is used. Use.
[0116]
The decoded value determination method shown in FIG. 29B is a method in which decoding is continued as much as possible even if an error is detected, and all available decoded values are used. A variable-length code includes a code having a feature that synchronization is automatically recovered when decoding is continued even when synchronization is lost. This is called self-synchronization recovery. In the case of a codeword having high self-synchronization recovery capability, a correct decoded value can be obtained by continuing decoding as shown in FIG. 29B without stopping the decoding process after error detection as shown in FIG. 29A. You can get more. In this case, however, there is a possibility that the decoded value includes a value decoded in error. In a system that allows such a possibility of erroneous decoding, it is possible to use the decoded value determination method of FIG.
[0117]
(Fourth embodiment)
Next, a variable-length encoding / decoding device according to the fourth embodiment of the present invention will be described with reference to FIG.
[0118]
The variable length coding / decoding apparatus according to the fourth embodiment is roughly composed of a codeword table creation unit 201, a coding unit 213, a transmission system or storage system 205, and a decoding unit 214. First, the functions of these units will be briefly described. The codeword table creation unit 201 creates a codeword table based on the occurrence probability of information symbols, and the codeword table in the coding unit 213 and the codeword table in the decoding unit 214. Codewords are sent to the forward codeword table and the reverse codeword table.
[0119]
The encoding unit 213 hierarchizes the information symbol given as input data, encodes it into variable length codes for each layer, multiplexes these variable length codes, and outputs them as encoded data to the transmission system or storage system 205 To do. The decoding unit 214 multiplexes and separates the encoded data input via the transmission system or the storage system 205, performs variable length decoding for each layer, and synthesizes the decoding results for each layer to restore the original information Play the symbol and output the playback data.
[0120]
Next, the detailed configuration and operation of each part of the fourth embodiment will be described. In the encoding unit 213, the information symbols that are input data are separated into layers of 1 to n (n is a natural number of 2 or more) according to the importance by the data layering unit 202. The separated information symbols of each layer i (i = 1, 2,..., N) are input to the variable length encoding unit 203-i prepared for each layer i and encoded. The encoded data of each layer i obtained by the variable length encoding unit 203-i is multiplexed by the multiplexing unit 204 and sent to the decoding unit 214 through the transmission system or the storage system 205.
[0121]
In the decoding unit 214, the encoded data input from the transmission system or the storage system 205 is separated into encoded data of each layer i by the multiplexing / separating unit 206. The separated encoded data of each layer is input to the variable length decoding unit 207-i prepared for each layer i and decoded. The decoding results of each layer i obtained by the variable length decoding unit 207-i are combined by the data combining unit 208 and output as reproduction data.
[0122]
The configuration of the codeword table creation unit 201 is the same as that of the codeword table creation unit 101 shown in FIG. 2 used in the first embodiment, and a code system that can be selected based on information on the occurrence probability of information symbols. The code having the smallest average code length is selected from among the codes, and the code word of the selected code is configured. The code word is a variable-length code (reversible code) that can be decoded both in the forward direction and in the reverse direction so that the code break can be recognized by a predetermined weight of the code word.
[0123]
Next, the configuration of the variable length encoding unit 203-i of each hierarchy i in FIG. 30 is as shown in FIG. 31, and the codeword table 102 and the code are similar to the encoding unit 113 shown in FIG. And the synchronization interval setting unit 104. That is, the information symbol input from the data hierarchization unit 202 in FIG. 30 is input to the encoder 103. The codeword table 102 stores information symbols created in advance by the codeword table creation unit 201 in FIG. 30 and codewords of variable length codes in association with each other.
[0124]
The encoder 103 selects and outputs a codeword corresponding to the information symbol input from the data hierarchization unit 202 from the codewords stored in the codeword table 102. The synchronization interval setting unit 104 collects the codewords selected by the encoder 103 for each synchronization interval, and further inserts a stuffing code that can be decoded in the forward direction or the reverse direction as necessary. Output encoded data. This encoded data is sent to the multiplexing unit 204 in FIG.
[0125]
On the other hand, the configuration of the variable length decoding unit 207-i of each hierarchy i in FIG. 30 is as shown in FIG. 32, and, similar to the decoding unit 114 shown in FIG. 107, a forward decoder 108, a backward decoder 109, a decoded value determination unit 110, a forward decoding table 111, and a backward decoding table 112. That is, in the variable length decoding unit 207-i, the synchronization interval detection unit 106 detects the synchronization interval of the encoded data from the encoded data input from the transmission system or the storage system 205, and the encoded data is stored in the buffer 107. accumulate. The forward decoder 108 starts decoding from the beginning of the encoded data stored in the buffer 107. The backward decoder 109 starts decoding from the end of the encoded data stored in the buffer 107.
[0126]
The forward decoder 108 detects an error when a bit pattern that does not exist in the forward codeword table 111 appears in the encoded data, or when encoded data having a bit number different from the bit number of the buffer 107 is decoded. Is determined. Similarly, in the reverse decoder 109, when a bit pattern that does not exist in the reverse codeword table 112 appears in the encoded data, and when encoded data having a bit number different from the bit number of the buffer 107 is decoded, Judged as error detection. The decoded value determination unit 110 determines a decoded value from the decoding result (forward decoding result) obtained by the forward decoder 108 and the decoding result (reverse decoded value) obtained by the backward decoder 109. The final decoding result is output. The decoding result of each layer i is sent to the data synthesis unit 208.
[0127]
FIG. 33 shows an example of data hierarchization in the data hierarchization unit 202 and multiplexing in the multiplexing unit 204. If there is syntax input data (information symbol string) that can be divided into n hierarchies as shown in FIG. 33 (a), the data hierarchizing unit 202 sets each hierarchy i as shown in FIG. 33 (b). The syntax is changed so that the input data is repeated, and data hierarchization is performed, and information symbols for each hierarchy i are sent to the variable length coding unit 203-i. In the multiplexing unit 204, as shown in FIG. 33 (c), encoded data in which a variable length encoding unit 203-i performs variable length encoding for each layer i and a synchronization interval is set by adding the synchronization code i. Are multiplexed and output.
[0128]
FIG. 34 shows an example of a decoded value determination method in the decoded value determination unit 110 in FIG. The variable length decoding unit 207-i for each layer i performs bi-directional decoding. If there is an error in the encoded data input to the variable length decoding unit of a certain layer, the influence affects the decoded values of the higher layer and the lower layer. The decoded value determination unit 110 in the variable length decoding unit 207-i exchanges information to determine the final decoded value.
[0129]
In the example of FIG. 34, if there is an error in the upper layer and there is a portion that cannot be decoded, a portion that cannot be decoded by the lower layer encoded data associated therewith is formed. Also, if there is an error in the lower layer encoded data and there is a portion that cannot be decoded, the decoded value of the upper layer encoded data corresponding thereto is changed. FIG. 34 is merely an example of the decoded value determination method, and the decoded value determination unit 110 in the variable length decoding unit 207-i of each layer i uses the decoding results of the upper layer and the lower layer to perform forward decoding. If an error is detected in at least one of the result and the backward decoding result, the decoding result of the forward decoding result and the backward decoding result which is estimated to be correct is used, and the forward decoding and the backward decoding are performed. Needless to say, any method may be used for the portion that could not be decoded as long as the decoding result is abandoned. In the fourth embodiment, the reversible code is used in both the forward and backward directions in all layers. However, the reversible code may be used in only some layers.
[0130]
(Fifth embodiment)
Next, the variable length encoding / decoding device described in the fourth embodiment is applied to the moving image multiplexing unit 703 and the moving image demultiplexing unit 707 of the moving image encoding / decoding system shown in FIG. A fifth embodiment will be described. In this case, the basic configuration and operation of the moving image encoding / decoding system are the same as those in the second embodiment, and variable length encoding and variable in the moving image multiplexing unit 703 and the moving image demultiplexing unit 707 are performed. The difference is that the long decoding process is performed for each layer.
[0131]
FIG. 35, FIG. 36, FIG. 37, and FIG. 38 show various examples of the syntax of the moving picture encoding method in the moving picture multiplexing unit 703 in the fifth embodiment. First, the syntax shown in FIG. 35 will be described. In this syntax, input information symbols that are encoded data are hierarchized into two layers, an upper layer and a lower layer, and a synchronization section is set for each by using synchronization codes RM and MM. ST at the end of the lower layer syntax is a stuffing code that can be decoded in the reverse direction as shown in FIG.
[0132]
As shown in FIG. 35 (a), the upper layer includes header information, an encoding mode for each macroblock, information on whether to encode each block, and the value of INTRA DC as mode information. These are described in variable-length codes that can be normally decoded in the forward direction, and then the motion vector information is variable-length encoded into a reversible code using the encoding tables shown in FIGS. 39 to 46, “VECTOR DIFFERENCES” represents a motion vector prediction value (difference value), “BIT NUMBER” represents a variable length code length, and “VLC CODE” represents a variable length code. .
[0133]
Here, the motion vector is obtained by one-dimensional prediction as shown in FIG. 47, and the predicted value is represented by a difference value. The motion vector prediction direction is indicated by an arrow in FIG. As shown in FIG. 47, there is one motion vector in a macro block (indicated by a large square in FIG. 47) and one motion vector for each luminance block (indicated by a small square in FIG. 47). There are cases where there are four, but one-dimensional prediction is performed for these, and the predicted values are described in variable-length codes in accordance with the encoding tables shown in FIGS. For motion vectors that cannot be obtained by prediction, the values of the motion vectors themselves are described by variable length codes according to the encoding tables shown in FIGS. The motion vector information at the end of FIG. 35A corresponds to this. On the other hand, the lower layer is DCT coefficient information excluding INTRA DC as shown in FIG. 35B, and is described by, for example, the reversible code described in FIGS. 14 to 20 and Japanese Patent Application No. 7-260383.
[0134]
Next, the syntax shown in FIG. 36 will be described. In this syntax, as shown in FIG. 36 (a), the upper layer is a variable that enables normal forward decoding by consolidating header information and mode information 1 representing the number of motion vectors for each macro coding at the head side. A long code is described, followed by a variable length code for motion vector information. The variable length code of the motion vector information is the same as the syntax of FIG. In the lower layer, as shown in FIG. 36 (b), mode information 2 indicating whether or not there is a DCT coefficient of each block and the INTRA DC are consolidated at the head side, and these can be variable-length capable of normal forward decoding. After that, the DCT coefficient information excluding INTRA DC is described in a reversible code as in the syntax of FIG.
[0135]
Next, the syntax shown in FIG. 37 will be described. In this syntax, only variable length codes that can be forward decoded are used in the upper layer. That is, as shown in FIG. 37 (a), the upper layer describes header information, mode information, and motion vector information with variable-length codes that can be forward decoded, and the lower layer as shown in FIG. 37 (b). The DCT coefficient information is described by a reversible code as in the syntax of FIG.
[0136]
The syntax shown in FIG. 38 also uses only variable length codes that can be forward decoded in the upper layer. That is, as shown in FIG. 38A, the upper layer describes header information, mode information 1 representing the number of motion vectors for each macro coding, and motion vector information in a variable length code that can be forward decoded. As shown in FIG. 38 (b), the lower layer is a variable length in which mode information 2 indicating whether or not there is a DCT coefficient of each block and INTRA DC are consolidated at the head, and these can be subjected to normal forward decoding. After that, the DCT coefficient information other than INTRA DC is described in a reversible code as in the syntax of FIG.
[0137]
The basic configuration of the moving picture multiplexing unit 703 and moving picture multiplexing / separating unit 707 in the fifth embodiment is as shown in FIGS. 13A and 13B, but the upper layer variable length encoder 901, the lower layer variable. The configurations of the long encoder 903, the upper layer variable length decoder 905, and the lower layer variable length decoder 906 are different from those of the second embodiment. That is, of the data encoded by the information source encoder 702, for example, upper layer data indicated by the syntax in FIG. 35A or FIG. 36A is variable by the upper layer variable length encoder 901. Long-coded and sent to the multiplexing unit 903. In addition, among the data encoded by the information source encoder 702, the lower-layer data indicated by the syntax in FIG. 35B or FIG. 36B is variable-length encoded by the variable-length encoder 903. And sent to the multiplexing unit 903. The multiplexing unit 903 multiplexes the upper layer encoded data and the lower layer encoded data, and sends the multiplexed data to the transmission buffer 704.
[0138]
In the fifth embodiment, the decoded value judgment unit 110 shown in FIG. 32 in the upper layer variable length decoder 905 and the lower layer variable length decoder 906 is obtained by the forward decoder 108. The decoding value is determined from the forward decoding result, and the final decoding result is output. The error determination in the forward decoder 108 and the backward decoder 109 of each layer is performed when a bit pattern that does not exist as a code word appears, and when an error is detected by a check bit or the like, the detection position is If no error is detected by the above-described determination method and the number of bits decoded does not match the number of bits of encoded data in the synchronization interval, the first position of decoding is set as the error detection position.
[0139]
Next, a decoded value determination method in the fifth embodiment will be described. FIG. 48 illustrates a decoded value determination method when the syntax of the video encoding scheme in the video multiplexing unit 703 is the syntax illustrated in FIG. In the syntax of FIG. 35, since reversible codes are used for motion vector information in the upper layer and DCT coefficient information in the lower layer, bidirectional decoding is possible in both the upper layer and the lower layer.
[0140]
In the upper layer, first, header information and mode information are decoded. When an error is found in the header information and the mode information and all cannot be decoded, the decoded values of the macroblocks in the synchronization section where the error has occurred are all Not Coded, and the previous screen is displayed as it is. If all the header information and mode information can be decoded, then the motion vector information is bidirectionally decoded. The portion that cannot be decoded with the motion vector information is also set as Not Coded. In the lower layer, the DCT coefficient information of the lower layer is used only for the macroblock that has been decoded to the motion vector. A macroblock in which DCT coefficient information is discarded due to an error is assumed to be Not Coded.
[0141]
FIG. 49 illustrates a decoded value determination method in the case where the moving image encoding scheme of the moving image multiplexing unit 703 has the syntax shown in FIG. According to the syntax shown in FIG. 36, header information and mode information 1 are first decoded in the upper layer. If an error is found in the header information or mode information 1 and all cannot be decoded, the decoded values of the macroblocks in the synchronization section where the error has occurred are all Not Coded, and the previous screen is displayed as it is. If the header information and the mode information 1 are all decoded, then the motion vector information is bidirectionally decoded. The portion that cannot be decoded with this motion vector information is also set as Not Coded. In the lower layer, the DCT coefficient information in the lower layer is used only for the macroblock that has been decoded up to the motion vector. A macroblock in which DCT coefficient information is discarded due to an error is assumed to be Not Coded.
[0142]
In the decoding value determination method shown in FIGS. 48 and 49, the decoding value is determined in units of macroblocks, but the decoding value may be determined in units of blocks or codewords. In the fifth embodiment, the case where the present invention is applied to variable length coding of motion vector information and DCT coefficient information has been described. However, the present invention can also be applied to other information symbols subjected to variable length coding. Needless to say.
[0143]
(Sixth embodiment)
Next, a variable-length encoding / decoding device according to a sixth embodiment of the present invention will be described. The basic configuration of the sixth embodiment is the same as the block diagram of FIG. 1 showing the configuration of the first embodiment, and will be described with reference to FIG.
[0144]
In the encoding unit 113, the input information symbol is input to the encoder 103. The codeword table 102 stores information symbols created in advance by the codeword table creation unit 101 in correspondence with codewords of variable length codes. The encoder 103 selects a code word corresponding to the input information symbol from the code words stored in the code word table 102. The synchronization interval setting unit collects the codewords selected by the encoder for each synchronization interval and outputs them as encoded data. The encoded data is sent to the decoding unit 111 through the transmission system or the storage system 104.
[0145]
In the decoding unit 114, the synchronization interval of the encoded data is detected by the synchronization interval detection unit 106 from the encoded data input from the transmission system or the storage system 105, and the encoded data is stored in the buffer 107. The forward decoder 109 starts decoding from the beginning of the encoded data stored in the buffer, and the backward decoder 111 starts decoding from the end of the encoded data stored in the buffer.
[0146]
In the forward decoder, if a bit pattern that does not exist in the forward codeword table appears in the encoded data, or if encoded data with a bit number different from the bit number of the buffer is decoded, it is determined that the error is detected. To do. Similarly, the backward decoder also detects errors when a bit pattern that does not exist in the backward codeword table appears in the encoded data, or when encoded data having a bit number different from the bit number of the buffer is decoded. Is determined.
[0147]
The decoded value determination unit determines a decoded value from the decoding result (referred to as the forward decoding result) obtained by the forward decoder and the decoded result (referred to as the backward decoded value) obtained from the backward decoder. The final decoding result is output.
[0148]
FIG. 50 is a block diagram of the codeword table creation unit 101. The occurrence probability of the information symbol is input to the probability model creating unit 23, a probability model is created, and the probability model is input to the code selecting unit 21, and the one having the smallest average code length is selected from the selectable code systems and selected. The result is sent to the codeword construction unit 22. The code word configuration unit 22 configures the code word of the code selected by the code word selection unit 21.
[0149]
The design method of the probability model in the sixth embodiment is as follows. In the sixth embodiment, a method for designing a more general and more efficient coding model for a probability distribution of various information sources will be described.
[0150]
First, the notation is shown.
θi Information source (i = 1, ..., n)
X Information source information symbol X = (x1, x2,..., Xm)
Frequency distribution of P (X | θi) θi
As an example, a frequency distribution obtained by encoding a test image using a plurality of test images as an information source may be considered.
[0151]
Here, the probability model Q (X) to be designed is obtained by weighting and averaging frequency distributions obtained from a plurality of information sources.
[0152]

At this time, the problem is how to obtain the weighting coefficient w (θi).
[0153]
The ideal code length L (X | θi) when the information source θi is encoded with Q (X) is:

It becomes. To minimize the ideal code length L (X | θi) on average for each information source

[0154]
This function is minimized when U (X) = Q (X).
w (θ1) = ... = w (θn) = 1 / n
It becomes.
[0155]
As another method for designing the probability model Q (X), a method for designing the probability model Q (X) assuming the worst information source will be described.
[0156]
The redundancy R (X | θi) when the information source θi is encoded by Q (X) is

It becomes. The weighted average S (X) for each information source of this redundancy is

Thus, the function indicates the mutual information amount of the event X and the event θ. To find the maximum value of this function,
S. Arimoto, "An algorithm for computing the capacity of arbitrary discrete memoryless channels," IEEE Trans. Inform. Theory, Vol. IT-18, pp. 14-20, 1972.
REBlahut, "Computation of channel capacity and rate-distortion functions," IEEE Trans. Inform. Theory, Vol. IT-18, pp. 460-473, 1972.
The Arimoto-Blahut algorithm shown in FIG. 2 is known, and w (θi) (i = 1,... N) in the worst case can be obtained by maximizing S (X) by these algorithms.
[0157]
As a design method of the probabilistic model Q (X), the method of minimizing the ideal code length on average and the method of assuming the worst on average are shown, but the two methods are combined to group information sources. A method may be applied in which several probability models are created by the former method and the latter method is used for the probability model.
[0158]
In the code selection unit 21 shown in FIG. 50, the probability model Q (X) created by the probability model creation unit 23 creates Q (Y) in which the information symbols X are sorted in descending order of probability, while the code construction unit 22 F (Z) is created by sorting the reversible codes created in step 1 in ascending order.

Is calculated and the one with the smallest value is selected, and a codeword table is created by associating information symbols with codewords.
[0159]
The code constructed by the codeword constructing unit 22 is reverse to the forward direction using the codeword weights shown in, for example, Japanese Patent Application Nos. 7-87777 and 7-260383 and the first and third embodiments. A variable length code that can also be decoded in the direction is used.
[0160]
(Seventh embodiment)
Next, a seventh embodiment of the present invention will be described. FIG. 51 is a conceptual diagram of a moving picture encoding / decoding apparatus incorporating a variable length encoding / decoding apparatus according to a seventh embodiment of the present invention. In the moving image encoder 709, the data encoded by the information source encoder 702 is subjected to variable length encoding, multiplexing, and the like in the moving image multiplexing unit 703, and smoothed in the transmission buffer 704. Then, it is sent to the transmission system or storage system 705 as encoded data. The encoding control unit 701 controls the information source encoder 702 and the moving image multiplexing unit 703 in consideration of the buffer amount of the transmission buffer 704. On the other hand, the moving image decoder 710 stores the encoded data from the transmission system or storage system 705 in the reception buffer 706, and then the moving image multiplexing / separating unit 707 demultiplexes the encoded data and performs variable length decoding. Is sent to the information source decoder 708, and finally the moving image is decoded.
[0161]
The syntax added in 7th Embodiment is demonstrated. FIG. 53 shows the syntax of the moving image encoding method in the moving image multiplexing unit 703 and the moving image demultiplexing / separating unit 707 in the seventh embodiment. The encoded data is hierarchized into an upper layer and a lower layer, and a synchronization section is set by using synchronization codes RM and MM, respectively. Further, ST in the lower hierarchy is a stuffing code that can be decoded from the reverse direction as shown in FIG. As shown in FIG. 63, the intra-frame encoded frame includes AC-DCT coefficient information including a part of mode information of macroblocks and INTRA DC as an upper layer, a part of mode information and AC-DCT coefficient information as a lower layer. A reversible code is applied. As shown in FIG. 64, the inter-frame encoded frame has a part of macro block mode information and motion vector information as an upper layer, a part of mode information, INTRA DC, and AC-DCT coefficient information as a lower layer. -A reversible code is applied to the DCT coefficient information.
[0162]
The upper layer and lower layer encoding and decoding shown in FIG. 53 are performed by the moving picture multiplexing unit shown in FIG. These configurations are substantially the same as those of the encoding device / decoding device according to the fifth embodiment. That is, the basic configuration of the moving picture multiplexing unit 703 and moving picture multiplexing / separating unit 707 in FIG. 51 is as shown in FIGS. 53 (a) and 53 (b), but the upper layer variable length encoder 801, lower layer The configurations of the variable length encoder 803, the upper layer variable length decoder 805, and the lower layer variable length decoder 806 are different from those of the second embodiment. That is, among the data encoded by the information source encoder 702, for example, the upper layer data indicated by the syntax in FIG. 53A is variable length encoded by the upper layer variable length encoder 801, and multiplexed. Sent to the conversion unit 803. In addition, among the data encoded by the information source encoder 702, lower-layer data indicated by the syntax in FIG. 53B is variable-length encoded by the variable-length encoder 803, and the multiplexing unit 803 Sent to. The multiplexing unit 803 multiplexes the upper layer encoded data and the lower layer encoded data, and sends the multiplexed data to the transmission buffer 704.
[0163]
In the seventh embodiment, the decoded value judgment unit 110 shown in FIG. 32 in the upper layer variable length decoder 805 and the lower layer variable length decoder 806 is obtained by the forward decoder 108. The decoding value is determined from the forward decoding result, and the final decoding result is output. The error determination in the forward decoder 108 and the backward decoder 109 of each layer is performed when a bit pattern that does not exist as a code word appears, and when an error is detected by a check bit or the like, the detection position is If no error is detected by the above-described determination method and the number of decoded bits does not match the number of bits of encoded data in the synchronization interval, the most processing position for decoding is set as the error detection position.
[0164]
(Eighth embodiment)
54 and 55 respectively show an example of the information source encoder 702 and the information source decoder 708 in the encoding / decoding device according to the eighth embodiment of the present invention. First, in FIG. 54, the input image signal is divided into macro blocks by the blocking circuit 401. The input image signal divided into macroblocks is input to the subtractor 402, and a difference from the predicted image signal is taken to generate a predicted residual signal. Either one of the prediction residual signal and the input image signal from the blocking circuit 401 is selected by the mode selection switch 403 and subjected to discrete cosine transform by a DCT (discrete cosine transform) circuit 404. The DCT coefficient data obtained by the DCT circuit is quantized by the quantization circuit 405. The quantized data is branched into two, and one is sent to the moving picture multiplexer 703 as DCT coefficient information. The other is that the inverse quantization circuit 406 and IDCT (Inverse Discrete Cosine Transform) circuit 407 sequentially perform the reverse processing of the quantization circuit 405 and DCT circuit 406, and then the adder 408 passes through the switch 410. A local decoded signal is generated by being added to the input predicted image signal. This local decoded signal is stored in the frame memory and input to the motion compensation circuit 411. The motion compensation circuit A010 performs motion compensation between the input image signal and the local decoded signal, obtains a motion vector, and generates a predicted image signal. The motion vector information obtained at this time is sent to the moving picture multiplexer 703. Based on the information from the motion compensation circuit 410, the mode selection circuit 412 determines what prediction method is used to encode each macroblock, and the result is the video multiplexing as the mode information together with the quantization parameter and the like. Sent to the container 703. In FIG. 54, the output encoded data is data in which three DCT encoding coefficients, motion vector information, and mode information are combined into one.
[0165]
Next, the configuration of the information source decoder 708 for decoding the data encoded by FIG. 54 will be described with reference to FIG. FIG. 55 shows an example of the information source decoder 708 corresponding to FIG. In FIG. 55, the information source decoder 708 receives the mode information, the motion vector information, the DCT coefficient information, etc. separated by the moving picture multiplexing / separating circuit 707.
[0166]
In the mode determination circuit 504 to which the mode information is input, if the mode information is INTRA, the mode changeover switch 505 is selected to be off, the DCT coefficient information is inversely quantized by the inverse quantization circuit 501, and the IDCT circuit 502 is inverted. A reproduced image signal is generated by performing a discrete cosine transform process. The reproduced image signal is stored as a reference image in the frame memory 506 and is output as a reproduced image signal. If the mode information is INTER, the mode changeover switch 505 is selected to be turned off, the DCT coefficient information is inversely quantized by the inverse quantization circuit 501, and the inverse discrete cosine transform processing is performed by the IDCT circuit 502, based on the motion vector information. Thus, the frame memory 504 compensates the motion of the reference image, and the adder 503 adds them to generate a reproduced image signal. The reproduced image signal is stored as a reference image in the frame memory 504 and is output as a reproduced image signal.
[0167]
(Ninth embodiment)
Next, as an application example of the present invention, an embodiment of an image transmitting / receiving apparatus to which the moving image encoding / decoding system according to the ninth embodiment in which the variable-length encoding / decoding apparatus of the present invention is incorporated is illustrated. 56 will be described. An image signal input from a camera 1002 provided in a personal computer (PC) 1001 is encoded by a moving image encoder 709 incorporated in the PC 1001. The encoded data output from the moving image encoder 709 is multiplexed with other audio and data information, transmitted wirelessly from the wireless device 1003, and received by the other wireless device 1004. The signal received by the wireless device 1004 is decomposed into encoded data and audio data of an image signal. Among these, the encoded data of the image signal is decoded by the moving image decoder 710 incorporated in the workstation (EWS) 1005 and displayed on the display of the EWS 1005.
[0168]
On the other hand, the image signal input from the camera 1006 provided in the EWS 1005 is encoded in the same manner as described above using the moving image encoder 709 incorporated in the EWS 1005. The encoded data is multiplexed with other voice and data information, transmitted wirelessly by the wireless device 1004, and received by the wireless device 1003. A signal received by the wireless device 1003 is decomposed into encoded data of an image signal and information of voice and data. Among these, the encoded data of the image signal is decoded by the moving image decoder 710 incorporated in the PC 1001 and displayed on the display of the PC 1001.
[0169]
(10th Embodiment)
The video encoding / decoding devices according to the first to ninth embodiments described above have taken the invention as a configuration as hardware, but the present invention has been developed to each of the above embodiments. It can also be understood as a recording medium for recording data or programs used in such an encoding / decoding device. Hereinafter, a recording medium on which data or a program according to the tenth embodiment is recorded will be described.
[0170]
In the information source encoder 702 shown in FIG. 51, a block-less scan is performed on a block of 8 × 8 DCT coefficients of a quantized word, and LAST (0: non-zero coefficient not the end of block, 1: end of block) RUN (the number of zero runs up to the non-zero coefficient) and LEVEL (quantized value of the coefficient) are obtained and sent to the moving picture multiplexing unit 703.
[0171]
The lower layer variable length encoder 802 in the moving picture multiplexing unit 703 supports the INDEX of the reversible code code word (VLC_CODE) to the RUN and LEVEL of the non-last coefficient of INTRA (intraframe coding) shown in FIG. 62, the INDEX table in which the RUN and LEVEL of the INTER (interframe coding) non-LAST coefficient shown in FIG. 61 are associated with the INDEX of the reversible codeword (VLC_CODE), and the INTRA shown in FIG. INDEX table that associates RUN and LEVEL of LAST coefficients common to INTER and INDEX of reversible code codeword (VLC-CODE), and codeword table that associates INDEX values and codewords shown in FIGS. 63 and 64 have.
[0172]
The operation of the lower layer variable length encoder 802 will be described with reference to the flowchart of FIG. First, based on the mode information, an INTRA non-LAST coefficient and a LAST coefficient codeword table are selected for INTER, and an INTER non-LAST coefficient and an LAST coefficient INDEX table are selected for INTER. (S101)
Next, when encoding using the codeword table, (RUN, LEVEL) is compared with the RUN maximum value R_max and the LEVEL maximum value L_max in the INDEX table to check whether they exist in the INDEX table. (S102)
If it exists, the INDEX table of FIGS. 60 to 62 is drawn to check whether the INDEX value is 0 (S104). If not, the INDEX code word of the code word table of FIG. 12 is output. To do. The last bit “s” of the codeword in the codeword table represents the LEVEL code. When “s” is “0”, the LEVEL code is positive, and when “s” is “1”, LEVEL. The sign of is negative (S105). When the INDEX value is 0 or out of the range, a coefficient that does not exist in the codeword table is continued with the ESCAPE code as shown in FIG. 66, the absolute values of RUN and LEVEL are fixed-length encoded, "00001" of two escapes is added to the head of this fixed-length encoding unit, and an escape code is also added to the end. The code word having the INDEX value of 0 in FIGS. 63 and 64 is an escape code, and the last bit “s” of VLC_CODE used as an escape (ESCAPE) code represents a LEVEL code, and “s” is “0”. LEVEL is positive, and if “1”, LEVEL is negative (S106).
[0173]
As described above, by using the INDEX table, even when the ESCAPE code is used, a code word can be output efficiently without searching. The lower layer variable length decoder 802 includes the decoded value table shown in FIGS. 68 and 69, and operates based on the INDEX value obtained as a result of decoding the variable length code.
[0174]
Next, the operation of the lower layer variable length decoder will be described with reference to the flowcharts of FIGS. FIG. 58 shows a decoding operation in the forward direction. First, variable length decoding is forward-decoded (S201). Next, it is checked whether or not the INDEX value obtained by decoding is 0 (S202). If it is other than 0, INTRA and INTER in the decoded value table of FIG. 68 are selected based on the mode information of the upper layer. Thus, the decoded values of LAST, RUN, and LEVEL are obtained (S203). If the INDEX value is 0, it is an ESAPE code, so the next fixed-length code is decoded to obtain the decoded values of LAST, RUN, and LEVEL (S204). Subsequently, the last ESCAPE code is decoded (S205). The sign of LEVEL is positive if the last bit of the code word is “0”, and negative if it is “1” (S206).
[0175]
FIG. 59 shows the decoding operation in the reverse direction. First, the sign of LEVEL is determined by the first bit of the code word. If it is “0”, it is positive, and if it is “1”, it is negative (S301). Next, the variable length code is reversely decoded (S302). It is checked whether the INDEX value obtained by backward decoding is 0 (S303). If it is other than 0, INTRA and INTER in the decoded value table of FIG. The decoded values of LAST, RUN, and LEVEL are obtained (S304). When the INDEX value is 0, since it is an ESCAPE code, the following fixed-length code is decoded to obtain decoded values of LEVEL, RUN, and LAST (S305). Subsequently, the head ESCAPE code is decoded (S306). As described above, by using the decoded value table, the storage amount can be saved even when the ESCAPE code is used, and the bidirectional decoding can be efficiently performed.
[0176]
A reinforcement example of the decoding process in the tenth embodiment will be described. That is, a decoding value determination method in the case where the syntax of the moving image encoding scheme in the moving image multiplexing unit 703 is the syntax shown in FIG.
[0177]
In the intra-coded frame, first, the mode information 1 and INTRADC of the upper layer are decoded. If an error is found and not all can be decoded, the decoded values of the macroblocks in the synchronization interval where the error has occurred are all Not Coded. If encoding is the first frame and there is no previous frame in time, the macroblock is filled with gray or a specific color.
[0178]
In the lower layer, when the mode information 2 cannot be decoded due to an error, all the encoded data in the lower layer is abandoned, and the decoded value in the upper layer is rewritten to be Not Coded or displayed only by INTRA DC. In addition, a macroblock in which the DCT coefficient is abandoned due to an error is displayed as Not Coded or displayed only by INTRA DC.
[0179]
Also, if there is a macroblock for which the AC-DCT coefficient is predicted in the INTRA mode macroblock, the variable length code can be decoded, but since the prediction is performed from the surrounding macroblock, the surrounding macroblock If the block cannot be decoded because it cannot be decoded, it is set to Not Coded.
[0180]
In the inter-frame coded frame, first, the mode information 1 and the motion vector of the upper layer are decoded. If an error is found and not all can be decoded, the decoded values of the macroblocks in the synchronization section where the error has occurred are all Not Coded. If all can be decoded, it is confirmed that the synchronization code MM exists. If not, all the decoded values of the macroblocks in the synchronization section are set to Not Coded.
[0181]
In the lower layer, when decoding is not performed due to an error in the mode information 2 and INTRA DC, the encoded data in the lower layer is discarded and displayed as Not Coded or displayed only with motion compensation from the previous frame (MC Not Coded) is rewritten with the decoded value of the upper layer. In addition, a macro block in which the DCT coefficient is abandoned due to an error is assumed to be MC Not Coded.
[0182]
Also, if there is a macroblock for which the AC-DCT coefficient is predicted in the INTRA mode macroblock, the variable length code can be decoded, but since the prediction is performed from the surrounding macroblock, the surrounding macroblock If the block cannot be decoded because it cannot be decoded, it is set to Not Coded.
[0183]
FIG. 70 shows a detailed decoded value determination method of the AC-DCT coefficient part of the inter-frame encoded frame.
[0184]
First, as shown in FIG. 70 (a), when the position of the macroblock in which an error is detected (error detection position) does not intersect between the forward decoding result and the backward decoding result, the macroblock in which no error is detected is detected. Only the decoding result is used as the decoding value, the decoding result at the two error detection positions is not used as the decoding value, and the previous frame is displayed as it is for the INTRA macroblock based on the decoding result of the higher layer mode information. Thus, for the INTRA macroblock, the decoding result of the upper layer is rewritten so that it is displayed with only motion compensation from the previous frame.
[0185]
Also, as shown in FIG. 70 (b), when the error detection positions of the forward decoding result and the backward decoding result intersect, the decoding values up to the macroblock in which the error is detected in the forward decoding result are The forward decoding result is used, and the subsequent decoding part uses the backward decoding result as a decoded value.
[0186]
In addition, the backward decoding result is prioritized, and the decoding value up to the macroblock in which an error is detected in the backward decoding result uses the backward decoding result, and the previous part decodes the forward decoding result. It may be used as a value.
[0187]
Furthermore, when an error is detected in both the forward decoding result and the backward decoding result in the same macroblock as shown in FIG. 70 (c), the decoded value of the macroblock at the error detection position is discarded and decoding is performed. Without being used as a value, based on the decoding result of the higher layer mode information, the previous frame is displayed as it is for the INTRA macroblock, and the motion compensation from the previous frame is displayed for the INTER macroblock. As described above, the decoding result of the upper layer is rewritten, and the decoding value for the subsequent macroblock uses the decoding result in the reverse direction.
[0188]
Also, as shown in FIG. 70 (d), when an error is detected by a stuffing code and decoding is not possible in the reverse direction, only forward decoding is performed. When an error is detected, macroblocks after the error detection position are Based on the decoding result of the higher layer mode information, for the INTRA macroblock, the previous frame is displayed as it is, and for the INTER macroblock, only the motion compensation from the previous frame is displayed. The decoding result is rewritten, and the decoding value for the subsequent macroblock uses the decoding result in the reverse direction.
[0189]
In the decoded value determination method of FIG. 70, the decoded value is determined in units of macroblocks. However, the determination may be performed in units of blocks or codewords.
[0190]
As described above, the variable-length encoding / decoding apparatus of the present invention requires a smaller calculation amount and storage amount than the conventional one, and is efficient.
[0191]
Specifically, the present invention can be applied to a moving image encoding / decoding device, and an efficient moving image encoding / decoding device can be provided with a smaller calculation amount and storage amount.
[0192]
【The invention's effect】
As described above, according to the present invention, variable length encoding / decoding capable of efficient encoding / decoding with less calculation amount and storage amount and decoding in both forward and backward directions is possible. An apparatus can be provided.
[0193]
In addition, according to the present invention, it is possible to reduce the useless bit pattern and increase the encoding efficiency, and also to decode in the forward direction and the reverse direction even when the synchronization interval is set in units of a fixed interval using the stuffing code. A variable-length encoding / decoding device can be provided.
[0194]
Further, according to the codeword construction method of the present invention, it is possible to deal with a wider probability distribution of information symbols, and it can be applied to coding with a large number of information symbols that could not be applied conventionally. Become. Specifically, the present invention can be applied to a moving image encoding / decoding system, and a moving image encoding / decoding system that is resistant to errors can be provided.
[0195]
Moreover, according to this codeword configuration method, coding efficiency is improved and code design with a high degree of freedom in the bit pattern of the codeword can be performed. Therefore, even when a synchronization interval is set using a synchronization code, The problem of pseudo-synchronization due to the coincidence of the bit pattern of the code and the code word can be avoided.
[0196]
Furthermore, according to the present invention, when determining the decoded value from the forward and backward decoding results, the decoded value of the encoded data is determined according to the error detection result when performing the forward and backward decoding. Thus, it is possible to effectively decode the reversible code against the transmission path error.
[0197]
In addition, according to the present invention, when decoding processing is performed by a bidirectional decoding means having an error detection function for a codeword of a variable length code, decoding processing of encoded data according to an error detection result in the forward decoding processing By switching between, it is possible to share the decoding processing in the forward direction and the backward direction, so that it is possible to effectively perform the decoding with respect to the transmission path error without significantly increasing the circuit scale and the calculation amount.
[0198]
Furthermore, according to the present invention, variable length coding is performed after the information symbol is hierarchized according to the importance on the encoding side, and encoded data for each layer is created for each synchronization interval, and the code for each layer is generated. By multiplexing the coded data, it is possible to perform decoding in both the forward and backward directions regardless of the information symbol syntax, and it is possible to perform variable-length coding / decoding that is more resistant to errors.
[Brief description of the drawings]
FIG. 1 is a block diagram showing a configuration of a variable-length encoding / decoding device according to a first embodiment of the present invention.
FIG. 2 is a block diagram showing a code word table creation unit in FIG. 1;
3 is a diagram for explaining a first code word configuration method in a code word configuration unit in FIG. 2; FIG.
FIG. 4 is a diagram illustrating forward and backward decoding trees created from codewords configured by the first code configuration method;
FIG. 5 is a diagram for explaining a second codeword configuration method in the codeword configuration unit in FIG. 2;
6 is a diagram for explaining a third code word construction method in the code word construction unit in FIG. 2; FIG.
7 is a diagram for explaining a synchronization interval setting method in a synchronization interval setting unit in FIG. 2;
FIG. 8 is a diagram showing a first example of a stuffing code.
FIG. 9 is a diagram illustrating a second example of the stuffing code.
10 is a diagram for explaining the operation of a decoded value determination unit in FIG. 2;
FIG. 11 is a block diagram showing a schematic configuration of a moving image encoding / decoding system according to a second embodiment of the present invention.
FIG. 12 is a diagram showing the syntax of encoded data in the moving image encoding / decoding system according to the second embodiment.
13 is a block diagram showing a configuration of a moving image multiplexing unit and a moving image multiplexing / separating unit in FIG.
FIG. 14 is a diagram showing a part of a codeword table of INTRA and INTER non-LAST coefficients in the second embodiment.
FIG. 15 is a diagram showing another part of a codeword table of INTRA and INTER non-LAST coefficients in the second embodiment.
FIG. 16 is a diagram showing still another part of a codeword table of INTRA and INTER non-LAST coefficients in the second embodiment.
FIG. 17 is a diagram showing a part of a codeword table of INTRA and INTER last coefficients in the second embodiment.
FIG. 18 is a diagram showing another part of a codeword table of INTRA and INTER last coefficients in the second embodiment;
FIG. 19 is a diagram showing a codeword table of escape codes in the second embodiment.
FIG. 20 is a diagram showing a coding method for codewords not in the codeword table in the second embodiment.
FIG. 21 is a diagram showing a case of a maximum zero run in the second embodiment.
FIG. 22 is a diagram for explaining the operation of a decoded value determination unit in the second embodiment.
FIG. 23 is a block diagram showing a configuration of a variable-length encoding / decoding device according to a third embodiment of the present invention.
24 is a diagram for explaining a first code word configuration method in the code word configuration unit in FIG. 23;
FIG. 25 is a diagram for explaining a second code word configuration method in the code word configuration unit in FIG. 23;
FIG. 26 is a diagram showing a configuration of a bidirectional codeword table in which a forward codeword table and a backward codeword table are shared.
FIG. 27 is a block diagram showing a configuration when an error detection unit is added to the decoder.
FIG. 28 is a diagram illustrating a decoding method according to the third embodiment.
FIG. 29 is a diagram for explaining a decoded value determination method of a decoded value determination unit in the third embodiment.
FIG. 30 is a block diagram showing a configuration of a variable-length encoding / decoding device according to a fourth embodiment of the present invention.
31 is a block diagram showing a configuration of a variable length encoding unit of one layer in FIG. 30. FIG.
32 is a block diagram showing a configuration of a variable length decoding unit of one layer in FIG. 30. FIG.
33 is a diagram showing an example of data hierarchization and multiplexing in the data hierarchization unit and the multiplexing unit in FIG. 30;
34 is a diagram showing an example of a decoded value determination method of a decoded value determination unit in FIG. 32. FIG.
35 is a moving picture multiplexing section and a moving picture multiplexing / demultiplexing section when the variable length encoding / decoding apparatus according to the fourth embodiment is incorporated in the moving picture encoding / decoding system of FIG. The figure which shows the 1st example of the syntax of an image coding system.
36 is a moving picture multiplexing unit and a moving picture demultiplexing / separating unit when the variable-length encoding / decoding device according to the fourth embodiment is incorporated in the moving image encoding / decoding system of FIG. The figure which shows the 2nd example of the syntax of an image coding system.
FIG. 37 shows moving images in the moving image multiplexing unit and the moving image demultiplexing unit when the variable length encoding / decoding device according to the fourth embodiment is incorporated in the moving image encoding / decoding system in FIG. The figure which shows the 3rd example of the syntax of an image coding system.
38 is a moving picture multiplexing section and moving picture multiplexing / demultiplexing section when the variable length coding / decoding apparatus according to the fourth embodiment is incorporated in the moving picture coding / decoding system of FIG. The figure which shows the 4th example of the syntax of an image coding system.
FIG. 39 is a diagram showing a motion vector encoding table according to the fourth embodiment;
FIG. 40 is a diagram showing a motion vector encoding table according to the fourth embodiment;
FIG. 41 is a diagram showing a motion vector encoding table according to the fourth embodiment;
FIG. 42 is a diagram showing a motion vector encoding table according to the fourth embodiment;
FIG. 43 is a diagram showing a motion vector encoding table according to the fourth embodiment;
FIG. 44 is a diagram showing a motion vector encoding table according to the fourth embodiment;
FIG. 45 is a diagram showing a motion vector encoding table according to the fourth embodiment;
FIG. 46 is a diagram showing a motion vector encoding table according to the fourth embodiment;
FIG. 47 is a diagram for explaining one-dimensional prediction of a motion vector in the fourth embodiment.
48 is a diagram for explaining a decoded value determination method in the case of the syntax in FIG. 35;
49 is a view for explaining a decoded value determination method in the case of the syntax in FIG. 36;
50 is a block diagram showing a detailed configuration of a codeword table creation unit in FIG. 1, which is an encoding / decoding device according to a sixth embodiment of the present invention.
FIG. 51 is a block diagram showing a schematic configuration of a moving image encoding / decoding system according to a seventh embodiment of the present invention.
52 is a block diagram showing a configuration of a moving image multiplexing unit and a moving image multiplexing / separating unit in FIG. 50. FIG.
FIG. 53 is a diagram illustrating an example of syntax of a moving image encoding scheme in a moving image multiplexing unit and a moving image demultiplexing unit according to the seventh embodiment.
FIG. 54 is a block diagram showing an example of an information source encoder according to the eighth embodiment of the present invention.
FIG. 55 is a block diagram showing an example of an information source decoder according to the eighth embodiment of the present invention.
FIG. 56 is a diagram showing an example of a system in which a variable-length encoding / decoding device according to a ninth embodiment of the present invention is incorporated.
FIG. 57 is a flowchart showing the operation of the variable length encoder of the tenth embodiment of the present invention.
FIG. 58 is a flowchart showing the operation of the forward variable length decoder according to the tenth embodiment.
FIG. 59 is a flowchart showing the operation of the backward variable length decoder according to the tenth embodiment.
FIG. 60 is a diagram showing an INDEX table of INTRA coefficients.
FIG. 61 is a diagram showing an INDEX table of INTER coefficients.
FIG. 62 is a diagram showing an INDEX table of LASR coefficients.
FIG. 63 shows a codeword table in the tenth embodiment.
FIG. 64 shows a codeword table in the tenth embodiment.
FIG. 65 is a diagram showing a RUN fixed-length codeword table in the tenth embodiment;
FIG. 66 is a diagram showing a LEVEL fixed-length codeword table according to the tenth embodiment;
Fig. 67 is a diagram showing a coding method for codewords not in the codeword table;
FIG. 68 is a decoded value table in the tenth embodiment.
FIG. 69 is a decoded value table in the tenth embodiment.
FIG. 70 is a diagram for explaining the operation of a decoded value determination unit;
Fig. 71 is a diagram illustrating a general decoding method of a reversible code.
FIG. 72 is an explanatory diagram of a normal variable length code.
FIG. 73 is an explanatory diagram of a conventional reversible code.
FIG. 74 is a diagram showing a conventional stuffing code.
FIG. 75 is a diagram for explaining the syntax of information symbols that cannot be subjected to conventional inverse decoding;
[Explanation of symbols]
51 Bidirectional codeword table
52 Address converter
53 Identification signal
61 Error detector
101 Codeword table creation unit
102 codeword table
103 Encoder
104 Synchronization section setting section
105, 205, 705 Transmission system or storage system
106 Synchronous section detector
107 buffers
108 Forward decoder
109 Reverse decoder
110 Decoded value determination unit
111 Forward codeword table
112 Reverse codeword table
113, 213 encoding unit
114, 121, 214 Decoding unit
122 Encoded data switcher
123 codeword table switcher
124,62 Decoder
201 Coding table creation unit
202 Data hierarchies
203 Variable length coding unit by layer
204 Multiplexer
206 Demultiplexer
207 Hierarchical decoding unit
208 Data composition part
401 Blocking circuit
402 Subtractor
403 Mode selection switch
404 DCT circuit
405 Quantization circuit
406 Inverse quantization circuit
407 IDCT circuit
408 Adder
409 frame memory
410 Motion Compensation Circuit
411 switch
412 Mode selection circuit
501 Inverse quantization circuit
502 IDCT circuit
503 Adder
504 Mode determination circuit
505 Mode selector switch
506 frame memory
701 Coding control unit
702 Information source encoder
703 video multiplexer
704 Transmission buffer
706 Receive buffer
707 Video demultiplexer
708 Information source decoder
709 video encoder
710 video decoder
901 Upper layer variable length encoder
902 Lower layer variable length encoder
903 Multiplexer
904 Demultiplexer
905 Upper layer variable length decoder
906 Lower layer variable length decoder
1001 Personal computer
1002 Workstation
1003, 1004 radio
1005, 1006 camera

Claims (6)

In a variable-length coding apparatus that assigns codewords having code lengths corresponding to the occurrence probability of information symbols to a plurality of information symbols, respectively, and outputs codewords corresponding to the input information symbols as coded data,
A hierarchization means for hierarchizing the input information symbols according to importance,
A plurality of codewords including codewords that can be decoded in both forward and reverse directions and that can be delimited by a predetermined weight of the codeword, both forward and backward In addition, a codeword formed by adding a second codeword decodable in both the forward and reverse directions to at least one of the beginning and end of the decodable first codeword corresponds to each information symbol. Codeword table stored and
Codeword selection means for selecting codewords corresponding to information symbols hierarchized by the hierarchization means from the codeword table;
Using the codeword selected by the codeword selection means to create encoded data for each layer for each predetermined synchronization interval, and multiplexing means for multiplexing the encoded data for each layer;
A variable-length coding apparatus comprising: In a variable-length coding apparatus that assigns codewords having code lengths corresponding to the occurrence probability of information symbols to a plurality of information symbols, respectively, and outputs codewords corresponding to the input information symbols as coded data,
A hierarchization means for hierarchizing the input information symbols according to importance,
A plurality of codewords including codewords that can be decoded in both forward and reverse directions and that can be delimited by a predetermined weight of the codeword, both forward and backward In addition, a codeword formed by adding a second codeword that can be decoded both in the forward direction and in the backward direction to at least one immediately before and after each bit of the first codeword that can also be decoded is an information symbol. Codeword tables stored corresponding to each of
Codeword selection means for selecting codewords corresponding to information symbols hierarchized by the hierarchization means from the codeword table;
Using the codeword selected by the codeword selection means to create encoded data for each layer for each predetermined synchronization interval, and multiplexing means for multiplexing the encoded data for each layer;
A variable-length coding apparatus comprising: A code word configured by adding a second code word that can be decoded in both the forward and backward directions to at least one of the head and the tail of the first code word is the head of the first code word. and at least one of the end, the variable length coding apparatus according to claim 1 in which the codeword of fixed length codes as the second code word is characterized by a code word that is inserted by a predetermined bit. A code word configured by adding a second code word that can be decoded in both the forward and backward directions to at least one of the head and the tail of the first code word is the head of the first code word. and at least one of the end, according to claim 1, wherein the second decodable codewords both backwards in the forward direction of the code word, characterized in that a code word that is inserted by a predetermined bit Variable length coding device. A codeword configured by adding a second codeword that can be decoded both in the forward direction and in the reverse direction to at least one of the first codeword immediately before and after each bit is the first codeword. The variable-length coding apparatus according to claim 2, wherein a code word of a fixed-length code as the second code word is inserted between each bit of the word by a predetermined number of bits.   A codeword configured by adding a second codeword that can be decoded both in the forward direction and in the reverse direction to at least one of the first codeword immediately before and after each bit is the first codeword. 3. The code word according to claim 2, wherein a code word that is decodable in both the forward and backward directions as the second code word is inserted between each bit of the word. Variable-length encoding device.

Images