JP3866624B2 - Moving picture encoding method, moving picture decoding method, moving picture encoding apparatus, and moving picture decoding apparatus - Google Patents

Description

予測モードの符号化方法は次の通りとする。まずｓｋｉｐかどうかを示すＣＯＤ情報（１ビット）があり，ｓｋｉｐの場合には１に設定し，それのみとする。ｓｋｉｐでない場合には０に設定し，続いて各ブロックの前方向予測か後方向予測か両方向予測かを示す情報（可変長４ビット）を符号化する。１の場合にはｄｉｒｅｃｔモードを示し，０１の場合には前方向予測を示し，００１の場合には後方向予測を示し，０００１の場合にはｂｉ−ｐｒｅｄｉｃｔｉｖｅモードを示すものとする。これを４ブロック分連続して符号化する。
【００７３】
このような前提で現マクロブロックは次のように符号化される。まず，Ｓｋｉｐ予測モード判定部３０３でｓｋｉｐマクロブロック内の各ブロックの予測モードを決定する。続いて再分割動き探索／符号化モード判定部３０２で各予測モードにおける評価値を求める。同時に前方向予測と後方向予測の場合には動きベクトルも求める。領域予測モード判定部３１５は，マクロブロック内の各ブロックの予測モードを決定し，さらにブロックがｓｋｉｐかどうかを決定する。決定した予測モードは，予測モード蓄積部３０４に蓄積する。
【００７４】
続いて前方向予測と判定された場合には，ＦｏｒｗａｒｄＭＣ部３０８にて予測画像を求め，後方向予測と判定された場合には，ＢａｃｋｗａｒｄＭＣ部３０９にて予測画像を求め，ｄｉｒｅｃｔモードと判定された場合には，ＤｉｒｅｃｔＭＣ部３０７にて予測画像を求め，ｂｉ−ｐｒｅｄｉｃｔｉｖｅモードと判定された場合には，Ｂｉ−ｐｒｅｄｉｃｔｉｖｅＭＣ部３１６にて予測画像を求める。ここでｓｋｉｐの場合には，式（２）で差分を０とした場合の動きベクトルを使用する。続いてｓｋｉｐではない場合には，予測誤差符号化／量子化部３１０で現画像と予測画像との間で予測誤差を求めて符号化し量子化する。
【００７５】
復号部３１１では，ｓｋｉｐではない場合には，予測誤差を逆量子化して復号し，ＦｏｒｗａｒｄＭＣ部３０８またはＢａｃｋｗａｒｄＭＣ部３０９またはＤｉｒｅｃｔＭＣ部３０７またはＢｉ−ｐｒｅｄｉｃｔｉｖｅＭＣ部３１６にて得られた予測画像を使って復号画像を求める。ｓｋｉｐの場合には，ＦｏｒｗａｒｄＭＣ部３０８またはＢａｃｋｗａｒｄＭＣ部３０９またはＤｉｒｅｃｔＭＣ部３０７またはＢｉ−ｐｒｅｄｉｃｔｉｖｅＭＣ部３１６にて得られた予測画像を復号画像とする。復号画像を参照画像メモリ３０６に蓄積する。前方向予測と後方向予測とｂｉ−ｐｒｅｄｉｃｔｉｖｅモードの場合には，動きベクトルを動きベクトル蓄積部３０５に蓄積する。
【００７６】
以上の手順をすべてのマクロブロックに対して繰り返し処理する。
【００７７】
次に復号装置の説明を行う。図７に，第２の実施の形態に係る画像復号装置の概要を示す。以下，図５に示す第２の実施の形態の画像符号化装置で得られた符号化データを復号する手順を説明する。
【００７８】
第２の実施の形態の画像復号装置は，符号化データ入力部４０１，予測モード復号部４０２，Ｓｋｉｐ予測モード判定部４０３，予測モード蓄積部４０４，動きベクトル蓄積部４０５，参照画像メモリ４０６，ＤｉｒｅｃｔＭＣ部４０７，ＦｏｒｗａｒｄＭＣ部４０８，ＢａｃｋｗａｒｄＭＣ部４０９，Ｂｉ−ｐｒｅｄｉｃｔｉｖｅＭＣ部４１６，予測誤差逆量子化／復号部４１０，スイッチ部４１２〜４１４を備える。
【００７９】
ＦｏｒｗａｒｄＭＣ部４０８では，前方向予測を行って予測画像を求め，ＢａｃｋｗａｒｄＭＣ部４０９では，後方向予測を行って予測画像を求め，ＤｉｒｅｃｔＭＣ部４０７では，ｄｉｒｅｃｔモード予測を行って予測画像を求め，Ｂｉ−ｐｒｅｄｉｃｔｉｖｅＭＣ部４１６では，ｂｉ−ｐｒｅｄｉｃｔｉｖｅモード予測を行って予測画像を求めるものとする。ＤｉｒｅｃｔＭＣ部４０７にてｄｉｒｅｃｔモードに使用する動きベクトルは，動きベクトル蓄積部４０５に蓄積されているものとする。
【００８０】
Ｓｋｉｐ予測モード判定部４０３では，予測モード蓄積部４０４に蓄積された左，左上，上，右上のマクロブロックの予測モードから，図６のフローに従ってｓｋｉｐマクロブロック内の全てのブロックの予測モードを決定する。
【００８１】
このような前提で現マクロブロックは次のように符号化される。まずＳｋｉｐ予測モード判定部４０３でｓｋｉｐマクロブロック内のブロックの予測モードを決定する。続いて予測モード復号部４０２で予測モードを復号する。まずＣＯＤ情報を復号する。１であった場合にはｓｋｉｐであり，０であった場合にはｓｋｉｐではない。ＣＯＤが０の場合に各ブロックにさらに最大４ビットの予測モード情報を復号する。１の場合にはｄｉｒｅｃｔモードであり，０１の場合には前方向予測であり，００１の場合には後方向予測であり，０００１の場合にはｂｉ−ｐｒｅｄｉｃｔｉｖｅ予測となる。ＣＯＤが１の場合には，Ｓｋｉｐ予測モード判定部４０３で決定した予測モードとなる。これを４ブロック分連続して復号する。
【００８２】
続いて予測モードが前方向予測の場合には，ＦｏｒｗａｒｄＭＣ部４０８にて予測画像を求め，後方向予測の場合には，ＢａｃｋｗａｒｄＭＣ部４０９にて予測画像を求め，ｄｉｒｅｃｔモードの場合には，ＤｉｒｅｃｔＭＣ部４０７にて予測画像を求め，ｂｉ−ｐｒｅｄｉｃｔｉｖｅモードの場合には，Ｂｉ−ｐｒｅｄｉｃｔｉｖｅＭＣ部４１６にて予測画像を求める。続いてｓｋｉｐではない場合には，予測誤差逆量子化／復号部４１０で予測誤差を逆量子化し復号し，予測誤差を使って復号画像を作成する。ｓｋｉｐの場合には，予測画像がそのまま復号画像となる。復号画像を参照画像メモリ４０６に蓄積する。前方向予測と後方向予測とｂｉ−ｐｒｅｄｉｃｔｉｖｅモードの場合には復号した動きベクトルを動きベクトル蓄積部４０５に蓄積する。復号した予測モードは予測モード蓄積部４０４に蓄積する。
【００８３】
以上の手順をすべてのマクロブロックに対して繰り返し処理する。
【００８４】
なお，以上の第１および第２の実施の形態では，Ｓｋｉｐ予測モード判定部１０３，３０３等で，図２または図６のフローに従ってｓｋｉｐ予測モードを求めたが，決定フローはこれに限るものではない。
【００８５】
さらに，画面内の位置によってＳｋｉｐ予測モード判定部１０３，３０３等の決定フローを変更してもよい。例えば，画面外周のマクロブロックまたはブロックの場合には，Ｓｋｉｐ予測モード判定部１０３，３０３等で，図２または図６のフローに従わず，ｄｉｒｅｃｔモードで予測することも好適である。
【００８６】
さらにまた，後方向予測における参照フレームの表示時刻が，現フレームよりも未来であるか過去であるかによって，Ｓｋｉｐ予測モード判定部１０３，３０３等の決定フローを変更してもよい。例えば後方向予測における参照フレームの表示時刻が現フレームよりも未来である場合には，図２または図６の決定フローに従い，後方向予測における参照フレームの表示時刻が現フレームよりも過去である場合には，常にｄｉｒｅｃｔモードにすることも好適である。
【００８７】
また，本実施の形態ではマクロブロックまたはブロック毎に予測誤差を符号化したが，複数フレーム分の画像情報をまとめて符号化しても良い。例えば，文献「石川ら，“マルチパラメータ動き補償を用いた動画像の３Ｄ／２Ｄハイブリッド符号化，”信学技報IE2001-76 ，pp.15-22，2001」のように複数フレーム分の画像情報をＤＣＴしたり，文献「J.-R.Ohm，“Three-dimentional subband coding with motion compensation ，”IEEE Trans. ，Image Process.，vol.3 ，pp.559-571，1994」のように複数フレーム分の画像情報をＷａｖｅｌｅｔ変換しても良い。
【００８８】
このように本実施の形態によれは，Ｂピクチャのｓｋｉｐマクロブロックのフレーム間予測方法を，隣接するマクロブロックまたはブロックの予測方法から適応的に変更することができる。
【００８９】
【発明の効果】
本発明によれば，Ｂピクチャのｓｋｉｐマクロブロックのフレーム間予測方法を，隣接するマクロブロックまたはブロックの予測方法から適応的に変更することができる。これにより，従来Ｂピクチャをはさむ３フレーム間で動きが一定しない場合に，ｄｉｒｅｃｔモードによる動きベクトルでは予測誤差を低減することができない場合においても，ｓｋｉｐマクロブロックのフレーム間予測モードをｄｉｒｅｃｔ以外に設定することが可能となり，隣接領域同士で画像内容に相関があることを利用して，符号化効率を向上することができる。
【図面の簡単な説明】
【図１】第１の実施の形態の画像符号化装置の例を示す図である。
【図２】第１の実施の形態のｓｋｉｐ予測モード決定フローの例を示す図である。
【図３】第１の実施の形態の動き探索／符号化モード判定部の動作フローの例を示す図である。
【図４】第１の実施の形態の画像復号装置の例を示す図である。
【図５】第２の実施の形態の画像符号化装置の例を示す図である。
【図６】第２の実施の形態のｓｋｉｐ予測モード決定フローの例を示す図である。
【図７】第２の実施の形態の画像復号装置の例を示す図である。
【図８】予測関係の例を説明するための図である。
【図９】ｄｉｒｅｃｔモードにおける動きベクトルの関係を説明するための図である。
【符号の説明】
１０１，３０１ 画像入力部
１０２ 動き探索／符号化モード判定部
２０２，４０２ 予測モード復号部
３０２ 再分割動き探索／符号化モード判定部
１０３，２０３，３０３，４０３ Ｓｋｉｐ予測モード判定部
１０４，２０４，３０４，４０４ 予測モード蓄積部
１０５，２０５，３０５，４０５ 動きベクトル蓄積部
１０６，２０６，３０６，４０６ 参照画像メモリ
１０７，２０７，３０７，４０７ ＤｉｒｅｃｔＭＣ部
１０８，２０８，３０８，４０８ ＦｏｒｗａｒｄＭＣ部
１０９，２０９，３０９，４０９ ＢａｃｋｗａｒｄＭＣ部
１１０，３１０ 予測誤差符号化／量子化部
２１０，４１０ 予測誤差逆量子化／復号部
１１１，３１１ 復号部
１１２〜１１４，２１２〜２１４，３１２〜３１４，４１２〜４１４ スイッチ部
３１５ 領域予測モード判定部
３１６，４１６ Ｂｉ−ｐｒｅｄｉｃｔｉｖｅＭＣ部[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a moving picture coding and decoding technique using a bi-directional inter-frame predictive coding method, and in particular, by adaptively changing the inter-frame prediction method of a skip macroblock of a B picture, The present invention relates to a moving picture coding method, a moving picture decoding method, a moving picture coding apparatus, and a moving picture decoding apparatus that can improve the above.
[0002]
[Prior art]
In general, in moving picture coding, inter-frame prediction coding is used in order to achieve high coding efficiency using temporal correlation. The frame encoding mode can be predicted from an I frame that is encoded without using correlation between frames, a P frame that is predicted from one previously encoded frame, and two previously encoded frames. There are B frames. In particular, the video encoding method H.264. In 263, decoded images for a plurality of frames can be stored in the reference image memory, and the reference image can be selected and predicted from the memory.
[0003]
The P frame is predicted from a previously encoded frame, while the B frame can be predicted from two previously encoded frames. Of the two frames used as reference frames, a method of predicting from a past frame is referred to as forward prediction, and a method of predicting from a future frame is referred to as backward prediction. The display time of the reference frame in the backward inter-frame prediction may be later than the current frame. In this case, after displaying the current frame, a reference frame for backward frame prediction is output. When prediction is performed from two frames in B frame (bidirectional inter-frame prediction), image information from two frames is interpolated to generate image information for one frame to generate a predicted image.
[0004]
FIG. 8A1 shows an example of the prediction relationship of moving images when the reference frame display time in the backward inter-frame prediction is in the future. When the encoding modes of the first frame to the seventh frame are encoded in the order of IBBPBBP, since there is a prediction relationship shown in FIG. 8 (A1), when encoding is actually performed, FIG. 8 (A2) As shown in FIG. 4, the frames are encoded in the order of (1) (4) (2) (3) (7) (5) (6).
[0005]
FIG. 8B1 shows an example of the prediction relationship of moving images when the reference frame display time in the backward inter-frame prediction is in the past. When the encoding modes of the first frame to the seventh frame are encoded in the order of IPBBPBB, since there is a prediction relationship shown in FIG. 8 (B1), when encoding is actually performed, FIG. 8 (B2) As shown in FIG. 4, the frames are encoded in the order of (1) (2) (4) (3) (5) (7) (6).
[0006]
Furthermore, these encoding modes can be designated for each area obtained by dividing the image, such as a macro block. Video coding method In H.263, there are only I macroblocks to be encoded using only intra-frame image information in the I frame. Within the P frame, there is a P macroblock predicted from one frame encoded in the past in addition to the I macroblock. It is also possible to change the image to be referenced for each macroblock. Within the B frame, there are B macroblocks in addition to I macroblocks and P macroblocks. As with the P macroblock, the B macroblock has a forward prediction mode or backward prediction mode in which prediction is performed from one frame encoded in the past, and a bidirectional prediction mode in which prediction is performed from two frames.
[0007]
The bidirectional prediction mode includes a bi-predictive mode that encodes a motion vector indicating a reference position in each frame and a direct mode that does not encode. In the direct mode, a forward motion vector and a backward motion vector are calculated from the motion vector of the macroblock at the same position in the future frame to be referred to.
[0008]
FIG. 9 shows a relationship diagram of motion vectors in the direct mode when the reference frame display time in the backward inter-frame prediction is in the future. In FIG. 9, P1 and P2 are already encoded frames, and Pc is an encoding target frame. Equation (1) shows a motion vector calculation formula when the motion vectors of the macroblocks between Pc, P1, and P2 are MVf and MVb, respectively. MVs is a motion vector of a macroblock between P1 and P2. In general, a time interval between frames P1 and Pc is applied to N1, a time interval between frames Pc and P2 is applied to N2, and a time interval between frames P1 and P2 is applied to D.
[0009]
MVf = MVs × N1 / D (1-1)
MVb = MVs.times.N2 / D (1-2)
In order to calculate the motion vector according to this equation (1), the motion vector is not encoded in the direct mode.
[0010]
In addition, when encoding a motion vector of a macroblock, there is a method of encoding a difference from the motion vectors of surrounding macroblocks. For example, the video encoding method H.264. In H.263, a motion vector predictor PMV is created from the motion vectors (MV1, MV2, MV3) of the left, upper, and upper right macroblocks using Equation (2), and the difference from the PMV is encoded. Note that Median (a, b, c) is an intermediate value among the values a, b, c.
[0011]
PMV = Median (MV1, MV2, MV3) (2)
According to this method, when there is little change from the motion vector of the surrounding macroblock, the motion vector can be encoded with higher encoding efficiency.
[0012]
A macroblock without encoded data is called a skip macroblock. Video coding method In H.263, the skip macroblock in the P frame has a motion vector of 0 and has no prediction error. Therefore, an image at the same position in the reference frame is set as a decoded image.
[0013]
The skip macroblock in the B frame obtains a motion vector by the same method as in the direct mode and has no prediction error. Therefore, an image obtained by interpolating the image information of the motion vector position calculated by the equation (1) is set as a decoded image.
[0014]
In general, for each macroblock, whether or not it is a skip macroblock, whether to perform intra coding, and coding mode information indicating an inter-frame prediction method are coded. H. In H.263, 1-bit COD encoding mode information indicating whether or not a skip macroblock is encoded. If not skip, MCBPC encoding mode information indicating whether to perform intra encoding or an inter-frame prediction method is encoded. To do. Therefore, in the case of a skip macroblock, information indicating an interframe prediction method, a motion vector, and a prediction error are not encoded. Video coding method In H.264, skip macroblocks do not encode COD coding mode information, but run-length encode the number of consecutive skip macroblocks.
[0015]
[Problems to be solved by the invention]
Since the direct mode of the B frame is obtained from the motion vector of the subsequent P frame as shown in FIG. 9, it is effective when there is a constant speed motion between three frames. That is, in the case of a moving image captured by the camera moving parallel at a constant speed, the prediction error can be greatly reduced by the motion vector in the direct mode. If there is no prediction error, it can be encoded as a skip macroblock. Since the skip macroblock is encoded with COD information or run-length encoded, the encoding efficiency of the entire B frame can be improved.
[0016]
However, when the motion is not constant between the three frames, the prediction error cannot be reduced with the motion vector in the direct mode. Therefore, skip macroblocks are not generated, and the encoding efficiency cannot be reduced.
[0017]
In view of the above points, the present invention has a skip macroblock (or a specific condition) even when the motion error in the direct mode cannot reduce the prediction error because the motion is not constant between the three frames. It is possible to set the skip macroblock inter-frame prediction mode to other than direct, and to improve the coding efficiency by utilizing the fact that there is a correlation in image contents between adjacent regions. And
[0018]
[Means for Solving the Problems]
In order to solve the above problems, the present invention uses the following method.
[0019]
  BookThe invention divides an input image into regions and performs bi-directional interframe predictive coding.OrForward interframe predictive coding schemeOr backward interframe predictive codingIn the moving picture coding method for coding picture information by using a frame, the current region is encoded with inter-frame prediction mode coded data, motion vector coded data, and prediction error coded data.WhenDo not haveskipInter-frame prediction of adjacent areas when coding as areasIf there are areas using the forward prediction mode and areas using the backward inter-frame prediction mode as modes, set the inter-frame prediction mode of the current area to bi-directional prediction mode and When there is no region that uses the backward prediction mode as the prediction mode, the interframe prediction mode of the current region is set to the forward prediction mode, and the forward prediction mode is used as the interframe prediction mode of the adjacent region. If there is no region, the inter-frame prediction mode of the current region is set to the backward prediction mode, and a prediction image is obtained from the reference image determined based on the set inter-frame prediction mode of the current region. Calculates the evaluation value of the encoding cost when the image is encoded, and determines whether to encode the current region as a skip region based on the evaluation valueIt is characterized by that.
[0021]
  BookAccording to the invention, even when the inter-frame prediction mode is not encoded as in the skip macroblock, the inter-frame prediction mode of the current region is changed according to a preset condition. be able to. For example, there are the following two conditions. These may be combined.
[Condition a] The most frequent mode among the inter-frame prediction modes in the four adjacent regions on the left, upper left, upper, and upper right is also applied to the current region.
[Condition b] When there are specific inter-frame prediction mode areas in the four adjacent areas on the left, upper left, upper, and upper right, the same mode is applied to the current area.
[0022]
Further, in the forward prediction, backward prediction, and bi-directional prediction, when a method of encoding a difference from the motion vector of the surrounding macroblock is applied, the 0 vector is applied as the motion vector of the skip macroblock. It is also possible to apply PMV calculated from equation (2).
[0023]
When there is a correlation in image content between adjacent regions, the encoding efficiency tends to be high in the same inter-frame prediction mode. Therefore, conventionally, when the motion is not constant between the three frames sandwiching the B picture, the inter-frame prediction mode of the skip macroblock is set to other than direct even when the prediction vector cannot reduce the prediction error. Thus, encoding efficiency can be improved.
[0024]
  Also, the aboveIn the present invention, during bi-directional inter-frame predictive encoding, a direct mode that does not encode a motion vector and a bi-predictive mode that encodes forward and backward motion vectors are distinguished, and an inter-frame prediction method for the current region is performed. decideYou may do it.
[0025]
  thisIf you doWhen the inter-frame prediction mode is not coded as in the skip macroblock, when the inter-frame prediction mode is obtained from the inter-frame prediction mode of the adjacent region, either the direct mode or the bi-predictive mode is used as the bidirectional prediction. Can be further distinguished. In the case of the direct mode, the motion vector can be obtained according to Expression (1), and in the case of the bi-predictive mode, the PMV obtained according to Expression (2) or the 0 vector can be applied.
[0026]
As a result, the inter-frame prediction mode can be changed more finely in the bidirectional prediction mode.
[0027]
  Also, the aboveIn the present invention, when encoding the current region as a region having no inter-frame prediction mode encoded data, motion vector encoded data, and prediction error encoded data, the current region is adjacent to each subdivision region further divided. Change the interframe prediction method for the current subdivision area according to whether the interframe prediction method for the subdivision area satisfies a preset condition.Such an implementation is also possible.
[0028]
  thisIf you doWhen the inter-frame prediction mode, the motion vector, and the prediction error are not encoded as in the skip macroblock, the inter-frame prediction mode can be changed for each further divided unit such as a block. As a result, the inter-frame prediction mode can be changed in finer units than the macroblock.
[0029]
  The image decoding method according to the present invention includes:Bidirectional interframe predictive coding or forward interframe predictive coding for each segmented imageOr backward interframe predictive codingIn the moving picture decoding method for decoding picture information by using the image, the current region is encoded with inter-frame prediction mode encoded data, motion vector encoded data, and prediction error encoded data.WhenDo not haveskipWhen decoding as an area, inter-frame prediction of adjacent areasIf there are areas using the forward prediction mode and areas using the backward inter-frame prediction mode as modes, set the inter-frame prediction mode of the current area to bi-directional prediction mode and When there is no region that uses the backward prediction mode as the prediction mode, the interframe prediction mode of the current region is set to the forward prediction mode, and the forward prediction mode is used as the interframe prediction mode of the adjacent region. If there is no region, the inter-frame prediction mode of the current region is set to the backward prediction mode, a reference image is determined based on the set inter-frame prediction mode of the current region, and the prediction image is determined from the reference image. Find and decode the current region imageIt is characterized by that.
[0031]
  Also, the aboveIn the present invention, at the time of bi-directional inter-frame prediction decoding, a direct mode in which a motion vector is not decoded and a bi-predictive mode in which a forward motion vector and a backward motion vector are decoded are distinguished, and an inter-frame prediction method for the current region is determined.You may do.
[0032]
  In addition, the aboveIn the present invention, when the current area is decoded as an area having no inter-frame prediction mode encoded data, motion vector encoded data, and prediction error encoded data, an adjacent re-division area is further divided for each re-divided area. Change the inter-frame prediction method for the sub-region of the current region according to whether the inter-frame prediction method for the sub-region satisfies the preset conditionImplementation is also possible.
[0033]
DETAILED DESCRIPTION OF THE INVENTION
Embodiments of the present invention will be described with reference to the drawings.
[0034]
In the first embodiment, B pictures are encoded by switching bi-directional prediction, forward prediction, and backward prediction for each macroblock. Here, the bi-directional prediction refers to the direct mode.
[0035]
First, the encoding device will be described. FIG. 1 shows an overview of an image coding apparatus according to the first embodiment. This image encoding apparatus includes an image input unit 101, a motion search / coding mode determination unit 102, a Skip prediction mode determination unit 103, a prediction mode storage unit 104, a motion vector storage unit 105, a reference image memory 106, a DirectMC unit 107, A ForwardMC unit 108, a BackwardMC unit 109, a prediction error encoding / quantization unit 110, a decoding unit 111, and switch units 112 to 114 are provided.
[0036]
The ForwardMC unit 108 performs forward prediction to obtain a prediction image, the BackwardMC unit 109 performs backward prediction to obtain a prediction image, and the DirectMC unit 107 performs bidirectional prediction to obtain a prediction image. It is assumed that motion vectors used for the direct mode in the DirectMC unit 107 are stored in the motion vector storage unit 105.
[0037]
In the skip prediction mode determination unit 103, the motion search / coding mode determination unit 102 converts the current macroblock from the prediction mode of the left, upper left, upper, and upper right macroblocks stored in the prediction mode storage unit 104 to the skip macro. Determine the prediction mode for block. The flow to determine is shown in FIG.
[0038]
First, it is determined whether there is a direct macroblock in any of the left, upper left, upper, and upper right of the current macroblock (step S1). If there is a direct macroblock, the skip macroblock prediction mode is determined to be the direct mode (step S2).
[0039]
If there are no direct macroblocks on the left, upper left, upper, and upper right, the number of forward prediction macroblocks Nf and the number of backward prediction macroblocks Nb in the left, upper left, upper, and upper right are counted (step S3). The number of predicted macroblocks Nf is compared with the number of backward predicted macroblocks Nb (step S4).
[0040]
If the number of macroblocks Nf and Nb are equal, the skip macroblock prediction mode is determined to be the direct mode (step S2).
[0041]
If the number of macroblocks Nf is greater than Nb, the skip macroblock prediction mode is determined to be the forward prediction mode (step S5).
[0042]
If the number of macroblocks Nf is smaller than Nb, the skip macroblock prediction mode is determined to be the backward prediction mode (step S6).
[0043]
FIG. 3 shows an operation flow of the motion search / coding mode determination unit 102 according to the present embodiment. The motion search / encoding mode determination unit 102 performs a motion search for forward prediction (step S10) and a motion search for backward prediction (step S11). In motion search, the prediction error and the code amount of the motion vector are evaluated using Equation (3), and the position with the smallest value is determined as the motion vector in the prediction method. It is assumed that the motion vector encodes the difference obtained using the equation (2).
[0044]
Further, the evaluation value Dmb of Expression (3) relating to the direct mode, the forward prediction, and the backward prediction is compared, and the prediction mode PM to be used when the value is the smallest is determined (step S12). When the determined prediction mode PM is the same as the prediction mode determined to be applied to the skip macroblock by the Skip prediction mode determination unit 103, the evaluation value obtained by Expression (3) satisfies Expression (4). (Step S13), if satisfied, the current macroblock is set to skip (step S14), otherwise, it is not set to skip (step S15).
[0045]
In Equation (3), A and B are values that vary depending on the quantization scale, p indicates a pixel, e [p] indicates a prediction error in the pixel p, and b (MV) indicates a motion vector. The code amount of MV is shown. The motion vector MV indicates a motion vector for forward prediction or a motion vector for backward prediction. THskip in equation (4) is a preset constant.
[0046]
Dmb = A * Σ | e [p] | + B * | b (MV) | (3)
(Where Σ is the sum for all pixels p in the macroblock)
Dmb ≦ THskip (4)
The encoding method for the prediction mode is as follows. First, there is COD information (1 bit) indicating whether it is a skip or not. If it is not skip, it is set to 0, and then information (2 bits) indicating forward prediction, backward prediction or bidirectional prediction is encoded. 01 indicates forward prediction, 10 indicates backward prediction, and 11 indicates bidirectional prediction.
[0047]
Under this assumption, the current macroblock is encoded as follows. First, the skip prediction mode determination unit 103 determines the prediction mode of the skip macroblock. Subsequently, the motion search / coding mode determination unit 102 determines the prediction mode. At the same time, in the case of forward prediction and backward prediction, a motion vector is also obtained. If the determined prediction mode is a skip macroblock prediction mode, it is determined whether or not the macroblock is to be skip. The determined prediction mode is stored in the prediction mode storage unit 104.
[0048]
Subsequently, the switch units 112 and 113 are switched. When it is determined that the forward prediction is performed, the ForwardMC unit 108 obtains a prediction image. When the backward prediction is determined, the BackwardMC unit 109 obtains a prediction image. If the direct mode is determined, the direct MC unit 107 obtains a predicted image. Here, in the case of skip, the motion vector when the difference is set to 0 in the equation (2) is used.
[0049]
Subsequently, when it is not skip, the prediction image is sent to the prediction error encoding / quantization unit 110 through the switch unit 113, and the prediction error encoding / quantization unit 110 sets the prediction error between the current image and the prediction image. Find and encode and quantize.
[0050]
If it is not skip, the decoding unit 111 dequantizes and decodes the prediction error, and obtains a decoded image using the prediction image obtained by the ForwardMC unit 108, BackwardMC unit 109, or DirectMC unit 107. In the case of skip, a predicted image obtained by the ForwardMC unit 108, the BackwardMC unit 109, or the DirectMC unit 107 is used as a decoded image. The decoded image is stored in the reference image memory 106. In the case of forward prediction and backward prediction, the motion vector is stored in the motion vector storage unit 105.
[0051]
The above procedure is repeated for all macroblocks.
[0052]
Next, the decoding device will be described. FIG. 4 shows an outline of the image decoding apparatus according to the first embodiment. A procedure for decoding the encoded data obtained in the embodiment of the image encoding device shown in FIG. 1 will be described.
[0053]
The image decoding apparatus includes an encoded data input unit 201, a prediction mode decoding unit 202, a Skip prediction mode determination unit 203, a prediction mode accumulation unit 204, a motion vector accumulation unit 205, a reference image memory 206, a DirectMC unit 207, a ForwardMC unit 208, A Backward MC unit 209, a prediction error inverse quantization / decoding unit 210, and switch units 212 to 214 are provided.
[0054]
The ForwardMC unit 208 performs forward prediction to obtain a prediction image, the BackwardMC unit 209 performs backward prediction to obtain a prediction image, and the DirectMC unit 207 performs bidirectional prediction to obtain a prediction image. It is assumed that the motion vector used in the direct mode in the DirectMC unit 207 is stored in the motion vector storage unit 205.
[0055]
The skip prediction mode determination unit 203 determines the prediction mode of the skip macroblock from the prediction modes of the left, upper left, upper, and upper right macroblocks accumulated in the prediction mode accumulation unit 204 according to the flow of FIG.
[0056]
Under this assumption, the current macroblock is decoded as follows. First, the skip prediction mode determination unit 203 determines the prediction mode of the skip macroblock. Subsequently, the prediction mode decoding unit 202 decodes the prediction mode. First, the COD information is decoded. When COD is 1, it is skip, and when it is 0, it is not skip. When COD is 0, 2 bits are further decoded. 01 is forward prediction, 10 is backward prediction, and 11 is bidirectional prediction. When COD is 1, the prediction mode determined by the Skip prediction mode determination unit 203 is set.
[0057]
Subsequently, when the prediction mode is forward prediction, a predicted image is obtained by the ForwardMC unit 208. When backward prediction is performed, a prediction image is obtained by the BackwardMC unit 209. When the prediction mode is direct mode, the DirectMC unit is obtained. In 207, a predicted image is obtained. Here, in the case of skip, the motion vector when the difference is set to 0 in the equation (2) is used.
[0058]
Subsequently, if it is not skip, the prediction error inverse quantization / decoding unit 210 dequantizes and decodes the prediction error, and creates a decoded image using the prediction error. In the case of skip, a predicted image becomes a decoded image as it is. The decoded image is stored in the reference image memory 206. In the case of forward prediction and backward prediction, the decoded motion vector is stored in the motion vector storage unit 205. The decoded prediction mode is stored in the prediction mode storage unit 204.
[0059]
The above procedure is repeated for all macroblocks.
[0060]
Next, a second embodiment will be described. In the second embodiment, B picture is encoded by switching forward prediction, backward prediction, direct mode, and bi-predictive mode for each block.
[0061]
First, the encoding device will be described. FIG. 5 shows an outline of an image coding apparatus according to the second embodiment. The image coding apparatus according to the second embodiment includes an image input unit 301, a subdivision motion search / coding mode determination unit 302, a Skip prediction mode determination unit 303, a prediction mode accumulation unit 304, and a region prediction mode determination unit 315. A motion vector storage unit 305, a reference image memory 306, a DirectMC unit 307, a Bi-predictive MC unit 316, a ForwardMC unit 308, a BackwardMC unit 309, a prediction error encoding / quantization unit 310, a decoding unit 311, and switch units 312 to 314 are provided. .
[0062]
The ForwardMC unit 308 performs forward prediction to obtain a prediction image, the BackwardMC unit 309 performs backward prediction to obtain a prediction image, and the DirectMC unit 307 performs direct mode prediction to obtain a prediction image, and Bi− The predictive MC unit 316 performs bi-predictive mode prediction to obtain a predicted image. It is assumed that the motion vector used for the direct mode in the DirectMC unit 307 is stored in the motion vector storage unit 305.
[0063]
The skip prediction mode determination unit 303 determines the prediction modes of all blocks in the skip macroblock from the prediction modes of the left, upper left, upper, and upper right blocks accumulated in the prediction mode accumulation unit 304. The flow to be determined is shown in FIG.
[0064]
First, it is determined whether there is a direct block in any of the left, upper left, upper, and upper right of the current block (step S20). If there is a direct block, the prediction mode of the block in the skip macroblock is determined to be the direct mode (step S25).
[0065]
When there are no direct blocks on the left, upper left, upper, and upper right, the number of forward prediction blocks Nf and the number of backward prediction blocks Nb in the left, upper left, upper, and upper right are counted (step S21), and the number of forward prediction blocks It is determined whether Nf and the backward prediction block number Nb are both 0 (step S22). If Nf = 0 and Nb = 0, the prediction mode of the block in the skip macroblock is determined as the direct mode (step S25).
[0066]
Unless Nf = 0 and Nb = 0, it is determined whether Nf> 0 and Nb> 0 (step S23). If Nf> 0 and Nb> 0, the prediction mode of the block in the skip macroblock is determined as the bi-predictive mode (step S25).
[0067]
Next, it is determined whether Nf> 0 and Nb = 0 (step S24). If Nf> 0 and Nb = 0, the prediction mode of the block in the skip macroblock is changed to the forward prediction mode. Determine (step S27).
[0068]
If Nf> 0 and Nb = 0, Nf = 0 and Nb> 0. In this case, the prediction mode of the block in the skip macroblock is determined as the backward prediction mode (step S28).
[0069]
The subdivision motion search / coding mode determination unit 302 performs a motion search for forward prediction and a motion search for backward prediction for each block. In motion search, the prediction error and the amount of code of the motion vector are evaluated by using Equation (5) described later, and the position where the value is the smallest is determined as the motion vector in the prediction method. In the bi-predictive mode, a prediction error is obtained using a motion vector obtained by motion search for forward prediction and backward prediction, and an evaluation value is obtained by Expression (6). The above processing is performed for all the blocks in the macro block.
[0070]
The region prediction mode determination unit 315 compares the evaluation values of the bi-predictive mode, the direct mode, the forward prediction, and the backward prediction obtained by the subdivision motion search / encoding mode determination unit 302 for each block. The prediction mode for the smallest value is obtained. The total evaluation value of the macroblock at this time is M1. Further, the evaluation value is also obtained by the equation (5) or the equation (6) for the prediction mode determined by the Skip prediction mode determination unit 303. The total evaluation value of the macroblocks at this time is M2.
[0071]
It is checked whether or not the evaluation value M1 and the evaluation value M2 satisfy the expression (7). If the evaluation value M1 and the evaluation value M2 satisfy the expression (7), the current macroblock is set to skip. In Equations (5) and (6), A, B, and C are values that change depending on the quantization scale, p indicates a pixel, and e [p] indicates a prediction error in the pixel p. , B (MV) indicate the code amount of the motion vector MV. The motion vector MVf indicates a forward prediction motion vector, the motion vector MVb indicates a backward prediction motion vector, and the motion vector MV indicates a forward prediction motion vector or a backward prediction motion vector. THskip in equation (7) is a preset constant. Σ represents the sum of all pixels p in the block.
[0072]

The prediction mode encoding method is as follows. First, there is COD information (1 bit) indicating whether it is a skip. In the case of a skip, it is set to 1 and only it. If it is not skip, it is set to 0, and then information (variable length 4 bits) indicating whether each block is forward prediction, backward prediction or bidirectional prediction is encoded. 1 indicates direct mode, 01 indicates forward prediction, 001 indicates backward prediction, and 0001 indicates bi-predictive mode. This is encoded continuously for 4 blocks.
[0073]
Under this assumption, the current macroblock is encoded as follows. First, the skip prediction mode determination unit 303 determines the prediction mode of each block in the skip macroblock. Subsequently, the re-division motion search / encoding mode determination unit 302 obtains an evaluation value in each prediction mode. At the same time, in the case of forward prediction and backward prediction, a motion vector is also obtained. The region prediction mode determination unit 315 determines the prediction mode of each block in the macroblock, and further determines whether the block is skip. The determined prediction mode is stored in the prediction mode storage unit 304.
[0074]
Subsequently, when the forward prediction is determined, the predicted MC unit 308 obtains a prediction image. When the backward prediction is determined, the Backward MC unit 309 obtains the prediction image, and the direct mode is determined. In this case, the DirectMC unit 307 obtains a predicted image, and when it is determined that the bi-predictive mode is selected, the Bi-predictive MC unit 316 obtains a predicted image. Here, in the case of skip, the motion vector when the difference is set to 0 in the equation (2) is used. Subsequently, if it is not skip, the prediction error encoding / quantization unit 310 obtains a prediction error between the current image and the prediction image, and encodes and quantizes the prediction error.
[0075]
In the decoding unit 311, if not skip, the prediction error is inversely quantized and decoded, and decoding is performed using the prediction image obtained by the ForwardMC unit 308, BackwardMC unit 309, DirectMC unit 307, or Bi-predictiveMC unit 316. Ask for an image. In the case of skip, a predicted image obtained by the ForwardMC unit 308, the BackwardMC unit 309, the DirectMC unit 307, or the Bi-predictiveMC unit 316 is used as a decoded image. The decoded image is stored in the reference image memory 306. In the case of forward prediction, backward prediction, and bi-predictive mode, motion vectors are stored in the motion vector storage unit 305.
[0076]
The above procedure is repeated for all macroblocks.
[0077]
Next, the decoding device will be described. FIG. 7 shows an outline of an image decoding apparatus according to the second embodiment. The procedure for decoding the encoded data obtained by the image encoding apparatus according to the second embodiment shown in FIG. 5 will be described below.
[0078]
The image decoding apparatus according to the second embodiment includes an encoded data input unit 401, a prediction mode decoding unit 402, a Skip prediction mode determination unit 403, a prediction mode storage unit 404, a motion vector storage unit 405, a reference image memory 406, and DirectMC. 407, ForwardMC unit 408, BackwardMC unit 409, Bi-predictiveMC unit 416, prediction error inverse quantization / decoding unit 410, and switch units 412 to 414.
[0079]
The ForwardMC unit 408 performs forward prediction to obtain a prediction image, the BackwardMC unit 409 performs backward prediction to obtain a prediction image, and the DirectMC unit 407 performs direct mode prediction to obtain a prediction image. The predictive MC unit 416 performs bi-predictive mode prediction to obtain a predicted image. It is assumed that motion vectors used in the direct mode in the DirectMC unit 407 are stored in the motion vector storage unit 405.
[0080]
The skip prediction mode determination unit 403 determines prediction modes of all blocks in the skip macroblock from the prediction modes of the left, upper left, upper, and upper right macroblocks accumulated in the prediction mode accumulation unit 404 according to the flow of FIG. To do.
[0081]
Under this assumption, the current macroblock is encoded as follows. First, the skip prediction mode determination unit 403 determines the prediction mode of blocks in the skip macroblock. Subsequently, the prediction mode decoding unit 402 decodes the prediction mode. First, the COD information is decoded. If it is 1, it is skip; if it is 0, it is not skip. When COD is 0, prediction mode information of a maximum of 4 bits is further decoded in each block. In case of 1, it is direct mode, in case of 01 it is forward prediction, in case of 001 it is backward prediction, and in case of 0001 it is bi-predictive prediction. When the COD is 1, the prediction mode determined by the Skip prediction mode determination unit 403 is used. This is decoded continuously for 4 blocks.
[0082]
Subsequently, when the prediction mode is forward prediction, a prediction image is obtained by the ForwardMC unit 408. When backward prediction is performed, a prediction image is obtained by the BackwardMC unit 409. When the prediction mode is direct mode, the DirectMC unit is obtained. A predicted image is obtained at 407, and in the case of the bi-predictive mode, a predicted image is obtained by the Bi-predictive MC unit 416. Subsequently, if it is not skip, the prediction error inverse quantization / decoding unit 410 dequantizes and decodes the prediction error, and creates a decoded image using the prediction error. In the case of skip, the predicted image becomes a decoded image as it is. The decoded image is stored in the reference image memory 406. In the case of forward prediction, backward prediction, and bi-predictive mode, the decoded motion vector is stored in the motion vector storage unit 405. The decoded prediction mode is stored in the prediction mode storage unit 404.
[0083]
The above procedure is repeated for all macroblocks.
[0084]
In the first and second embodiments described above, the skip prediction mode is determined by the skip prediction mode determination units 103 and 303 according to the flow of FIG. 2 or FIG. 6, but the determination flow is not limited to this. Absent.
[0085]
Furthermore, the determination flow of the skip prediction mode determination units 103 and 303 may be changed according to the position in the screen. For example, in the case of a macro block or a block on the outer periphery of the screen, it is also preferable to perform prediction in the direct mode without skipping the flow of FIG. 2 or FIG.
[0086]
Furthermore, the determination flow of the skip prediction mode determination units 103 and 303 may be changed depending on whether the display time of the reference frame in the backward prediction is in the future or in the past than the current frame. For example, when the display time of the reference frame in the backward prediction is later than the current frame, the display time of the reference frame in the backward prediction is earlier than the current frame according to the determination flow of FIG. 2 or FIG. It is also preferable to always set the direct mode.
[0087]
In this embodiment, the prediction error is encoded for each macroblock or block, but image information for a plurality of frames may be encoded together. For example, as in the document “Ishikawa et al.,“ 3D / 2D hybrid encoding of moving images using multi-parameter motion compensation, ”IEICE Technical Report IE2001-76, pp.15-22, 2001” Information can be DCTed, and multiple documents such as “J.-R.Ohm,“ Three-dimentional subband coding with motion compensation ”, IEEE Trans., Image Process., Vol.3, pp.559-571, 1994”. Image information for frames may be wavelet transformed.
[0088]
As described above, according to the present embodiment, the inter-frame prediction method for a skip macroblock of a B picture can be adaptively changed from the prediction method for an adjacent macroblock or block.
[0089]
【The invention's effect】
According to the present invention, the inter-frame prediction method for a skip macroblock of a B picture can be adaptively changed from the prediction method for an adjacent macroblock or block. As a result, when the motion is not constant between the three frames sandwiching the conventional B picture, even when the motion vector in the direct mode cannot reduce the prediction error, the interframe prediction mode of the skip macroblock is set to other than direct. Coding efficiency can be improved by using the fact that there is a correlation in image contents between adjacent regions.
[Brief description of the drawings]
FIG. 1 is a diagram illustrating an example of an image encoding device according to a first embodiment.
FIG. 2 is a diagram illustrating an example of a skip prediction mode determination flow according to the first embodiment;
FIG. 3 is a diagram illustrating an example of an operation flow of a motion search / coding mode determination unit according to the first embodiment;
FIG. 4 is a diagram illustrating an example of an image decoding device according to the first embodiment.
FIG. 5 is a diagram illustrating an example of an image encoding device according to a second embodiment;
FIG. 6 is a diagram illustrating an example of a skip prediction mode determination flow according to the second embodiment;
FIG. 7 is a diagram illustrating an example of an image decoding device according to a second embodiment.
FIG. 8 is a diagram for explaining an example of a prediction relationship.
FIG. 9 is a diagram for explaining the relationship of motion vectors in the direct mode.
[Explanation of symbols]
101, 301 Image input unit
102 Motion search / coding mode determination unit
202, 402 Prediction mode decoding unit
302 Subdivision motion search / coding mode determination unit
103, 203, 303, 403 Skip prediction mode determination unit
104, 204, 304, 404 Prediction mode storage unit
105, 205, 305, 405 Motion vector storage unit
106, 206, 306, 406 Reference image memory
107, 207, 307, 407 DirectMC part
108, 208, 308, 408 ForwardMC part
109, 209, 309, 409 BackwardMC part
110, 310 Prediction error encoding / quantization unit
210, 410 Prediction error inverse quantization / decoding unit
111,311 decoding unit
112-114, 212-214, 312-314, 412-414 Switch part
315 Region prediction mode determination unit
316, 416 Bi-predictive MC part

Claims (4)

In a moving image coding method for dividing an input image into regions and coding image information using a bidirectional inter-frame predictive coding method, a forward inter-frame predictive coding method, or a backward inter-frame predictive coding method,
When coding the current area as the skip area having no inter-frame prediction mode encoded data and the motion vector coding data and the prediction error coded data,
If there is an area that uses the forward prediction mode and an area that uses the backward interframe prediction method as the interframe prediction mode of the adjacent area, the interframe prediction mode of the current area is set to the bidirectional prediction mode. ,
If there is no region that uses the backward prediction mode as the inter-frame prediction mode of the adjacent region, set the inter-frame prediction mode of the current region to the forward prediction mode,
If there is no region that uses the forward prediction mode as the inter-frame prediction mode of the adjacent region, set the inter-frame prediction mode of the current region to the backward prediction mode,
An evaluation value of a coding cost when a prediction image is obtained from a reference image determined based on the set inter-frame prediction mode of the set current area and an image of the current area is encoded is calculated. A moving picture coding method characterized by determining whether or not to code a region as a skip region . In a moving picture decoding method for decoding image information using a bidirectional inter-frame predictive encoding scheme, a forward inter-frame predictive encoding scheme, or a backward inter-frame predictive encoding scheme for each region obtained by dividing an image,
When decoding the current area as the skip area having no inter-frame prediction mode encoded data and the motion vector coding data and the prediction error coded data,
If there is an area that uses the forward prediction mode and an area that uses the backward interframe prediction method as the interframe prediction mode of the adjacent area, the interframe prediction mode of the current area is set to the bidirectional prediction mode. ,
If there is no region that uses the backward prediction mode as the inter-frame prediction mode of the adjacent region, set the inter-frame prediction mode of the current region to the forward prediction mode,
If there is no region that uses the forward prediction mode as the inter-frame prediction mode of the adjacent region, set the inter-frame prediction mode of the current region to the backward prediction mode,
A moving picture decoding method comprising: determining a reference picture based on the set inter-frame prediction mode of the current area, obtaining a predicted picture from the reference picture, and decoding the picture of the current area . In a video encoding apparatus that divides an input image into regions and encodes image information using a bidirectional inter-frame predictive encoding scheme, a forward inter-frame predictive encoding scheme, or a backward inter-frame predictive encoding scheme,
An image input unit for inputting an image for each area;
A prediction mode storage unit for storing the inter-frame prediction mode of the already encoded region;
When coding the current area as the skip area with no inter-frame prediction mode encoded data and the motion vector coding data and the prediction error coding data, with reference to the prediction mode stored in the prediction mode storage unit, adjacent If there are an area using the forward prediction mode and an area using the backward prediction mode as the inter-frame prediction mode of the area, the inter-frame prediction mode of the current area is set to the bidirectional prediction mode, and the adjacent area If there is no region that uses the backward prediction mode as the inter-frame prediction mode of the current region, the inter-frame prediction mode of the current region is set to the forward prediction mode, and the forward prediction mode is set as the inter-frame prediction mode of the adjacent region. If there is no region in which use is skip prediction mode determining unit for setting the inter-frame prediction mode for the current region in the backward prediction mode ,
A motion search is performed between the input image and the reference image, the prediction error in the forward prediction mode, the backward prediction mode, or the bidirectional prediction mode is obtained, and the prediction error is used to determine which prediction mode is used for encoding. , Determining a motion vector to be encoded according to the determined prediction mode, and regarding the inter-frame prediction mode set by the skip prediction mode determination unit, a prediction image is determined from a reference image determined based on the inter-frame prediction mode. A motion search / encoding mode determination unit that determines whether to encode the current region as the skip region based on an evaluation value of the encoding cost when the current region image is encoded
A forward motion compensation unit for creating a prediction image used for motion compensation in the forward prediction mode;
A backward motion compensation unit for creating a prediction image used for motion compensation in the backward prediction mode;
A bidirectional motion compensator for creating a prediction image used for motion compensation in the bidirectional prediction mode;
A prediction error encoding / quantization unit that obtains a prediction error, encodes it, and quantizes it;
A decoding unit that decodes the quantized encoded data to create a decoded image;
A reference image memory for storing decoded images;
A motion vector storage unit for storing motion vectors;
A moving picture encoding apparatus comprising: In a video decoding device that decodes image information using a bidirectional inter-frame predictive coding method, a forward inter-frame predictive coding method, or a backward inter-frame predictive coding method for each region obtained by dividing an image,
Information indicating whether or not the current area is encoded as a skip area having no inter-frame prediction mode encoded data, motion vector encoded data, and prediction error encoded data, and a frame when the current area is not the skip area and prediction mode decoding unit that decodes the between prediction mode,
A prediction mode storage unit for storing the inter-frame prediction mode of the already decoded area;
When the current region is the skip region, the inter-frame prediction mode stored in the prediction mode storage unit is referred to, and the region using the forward prediction mode and the backward prediction mode are used as the inter-frame prediction mode of the adjacent region. If there are and a region in which use is to set the inter-frame prediction mode of the current area in the bidirectional prediction mode, if there is no area that uses the backward prediction mode as the inter-frame prediction modes of the adjacent region, set the inter-frame prediction mode for the current region in the forward prediction mode, if there is no area that uses a forward prediction mode as the inter-frame prediction modes of the adjacent regions after the inter-frame prediction mode of the current region A skip prediction mode determination unit for setting the direction prediction mode ;
When the inter-frame prediction mode decoded by the prediction mode decoding unit or the inter-frame prediction mode set by the skip prediction mode determination unit is the forward prediction mode, a prediction image used for motion compensation is generated in the forward prediction mode A forward motion compensation unit;
When the inter-frame prediction mode decoded by the prediction mode decoding unit or the inter-frame prediction mode set by the skip prediction mode determination unit is the backward prediction mode, a prediction image used for motion compensation is generated in the backward prediction mode A backward motion compensation unit;
Bidirectional motion for creating a prediction image used for motion compensation in bidirectional prediction mode when the interframe prediction mode decoded by the prediction mode decoding unit or the interframe prediction mode set by the skip prediction mode determination unit is a bidirectional prediction mode A compensation section;
A prediction error dequantization / decoding unit for dequantizing and decoding the prediction error;
A reference image memory for storing decoded images;
A motion vector storage unit for storing motion vectors;
A moving picture decoding apparatus comprising:

Images