WO2007074543A1

WO2007074543A1 - Moving picture image decoding device and moving picture image coding device

Info

Publication number: WO2007074543A1
Application number: PCT/JP2006/303999
Authority: WO
Inventors: Tomoyuki Yamamoto; Maki Takahashi
Original assignee: Sharp Kabushiki Kaisha
Priority date: 2005-12-27
Filing date: 2006-03-02
Publication date: 2007-07-05
Also published as: JPWO2007074543A1; JP5020829B2; JP2011160468A

Abstract

A moving picture image decoding device and a moving picture image coding device include a prediction vector deriving means in which not less than two reference images are candidates for a standard image. When one of standard image candidates is selected as the standard image and no standard moving vector exists because a standard region is intra-coded, a separate standard image candidate is selected as the standard image, so that possibility to derive a prediction vector matched with a real image movement can be enhanced. Thus, even though a standard region on a backward reference image is intra-coded, other prediction vectors than a zero vector can be derived, and a moving picture image decoding device and a moving picture image coding device are prevented from deterioration in prediction efficiency.

Description

Specification

Moving picture decoding apparatus and moving picture encoding apparatus

Technical field

The present invention relates to a moving image encoding device and a moving image decoding device that use a plurality of reference images for motion compensation.

Background art

As a conventional technique, a moving picture decoding apparatus using the MPEG-4 AVC method (Non-patent Document 1: ISO / lEC 14496-10) will be described below with reference to the block diagram shown in FIG.

[0003] The moving picture decoding apparatus 3 includes a variable length code decoding unit 100, a motion vector decoding unit (motion vector decoding means) 101, a prediction vector deriving unit 102, a buffer memory 103, an image decoding unit 104, and a predicted image deriving unit. (Predicted image deriving means) 105 is configured.

[0004] The variable-length code decoding unit 100 performs variable-length decoding on input encoded data, and decodes code information such as a prediction scheme, prediction residual data, difference margin, and time information. The motion vector decoding unit 101 decodes a motion vector from the prediction vector and the difference vector. The prediction vector deriving unit 102 derives a prediction vector using the decoded motion vector based on the prediction method. The buffer memory 103 temporarily records motion vectors, images, time information, and the like. The image decoding unit 104 decodes an image from the prediction method, the prediction residual data, and the predicted image. The predicted image deriving unit 105 derives a predicted image by motion compensation using the motion vector and the reference image.

[0005] <Outline of decryption procedure using conventional technology>

The outline of the decoding process in the moving picture decoding apparatus using the prior art will be described below with reference to the flowchart of the decoding process procedure shown in FIG.

[0006] First, the variable-length code decoding unit 100 performs variable-length decoding on code data input from the outside of the video decoding device 3 (step 10, hereinafter abbreviated as S10). The output of the variable-length code decoding unit 100 is encoded information such as a prediction method, prediction residual data, a difference vector, and time information. The prediction method is output to the prediction vector deriving unit 102 and the image decoding unit 104. The prediction residual data is output to the image decoding unit 104, the difference vector is output to the motion vector decoding unit 101, and the time information is output to the image decoding unit 104.

Next, the prediction vector deriving unit 102 derives a prediction vector according to the prediction method input from the variable length code decoding unit 100 using the motion vector recorded in the buffer memory 103 ( S20). The prediction vector deriving unit 102 outputs the derived prediction vector to the motion vector decoding unit 101 and the buffer memory 103. Details of the operation of the prediction vector deriving unit 102 will be described later.

[0008] Next, the motion vector decoding unit 101 adds the difference vector input from the variable length code decoding unit 100 to the prediction vector input from the prediction vector deriving unit 102, and outputs it as a motion vector ( S30). The output motion vector is output to the notch memory 103 and recorded.

Next, the predicted image deriving unit 105 reads the reference image recorded in the buffer memory 103. The predicted image derivation unit 105 performs motion compensation prediction using the motion vector input from the motion vector decoding unit 101 via the buffer memory 103 and the read reference image, and derives a predicted image. The derived predicted image is output to the image decoding unit 104 (S40).

Next, the image decoding unit 104 is input from the prediction image input from the prediction image deriving unit 105 and the variable length code decoding unit 100 according to the prediction scheme input from the variable length code decoding unit 100. The image is decoded based on the predicted residual data (S50). The decoded image and time information regarding the display timing of the image are output to the buffer memory 103 and recorded.

[0011] Next, the image recorded in the buffer memory 103 in S50 is output to a moving image display device (not shown) at the time indicated by the time information (S60).

[0012] As described above, in the conventional technique (MPEG-4 AVC method), decoding using motion compensation prediction is performed.

[0013] <About temporal direct prediction>

In the MPEG-4 AVC method, a highly efficient motion compensation prediction method called temporal direct prediction can be used as one of prediction methods.

Next, the conceptual diagram shown in FIG. 11 for temporal direct prediction of the MPEG-4 AVC method. Will be described. In the following description, unless otherwise specified, there are two reference images, one for forward reference and one for backward reference.

[0015] Temporal direct prediction is to derive a predicted image of the target image P as shown in FIG.

cur

In this case, two reference images, a forward reference image P and a backward reference image P, are used.

rO, 0 rl, 0

This is a motion compensation prediction method.

[0016] In the following, the processing target area (white circle in FIG. 11) on the target image P is referred to as target area A.

cur cur In the conceptual diagram of FIG. 11 and the conceptual diagrams shown in FIGS. 4, 5, 6, and 14, the left and right directions in the figure represent the display time for displaying moving images, and the vertical bars represent the images. The vertical direction in the figure represents the position of the area in each image.

[0017] In motion compensation in temporal direct prediction, the forward prediction vector mv of the target region A

cur

L0 and backward prediction vector mvLl are used. The forward prediction vector mvLO and the backward prediction vector mvLl are spatially the same as the target area A on the image P defined as the “reference image”.

col cur

Calculation is based on the motion vector mvCol of region B at one position (double circle in Fig. 11). Na

col

Region B is called the “reference region”, and the motion vector mvCol is called the “reference motion vector”.

丄

I will do it. Also, the area pointed to by the reference motion vector mvCol is area B, and area B

colref colref shall be on image P.

coirei

In the temporal direct prediction of the MPEG-4 AVC method, as shown in FIG. 11, the back reference image P and the base image P are used. Also, especially when there are two reference images, forward reference rl, 0 col

Image p I Image ρ.

rO, 0 coiref

The forward prediction vector mvLO and the backward prediction vector mvLl are calculated using the reference motion vector mvC ol according to the following equation.

mvLl = (tb-td) / tbb X mvCol (2)

In Equation (1) and Equation (2), tb is the forward reference image P and the target image P.

rO, 0 cur Display time interval, td is the time interval between forward reference image P and backward reference image P

r0, 0 rl, 0

Tbb represents the time interval between the image P and the reference image P. Time intervals tb, td, and tb

coirei col

Each value of b is calculated using time information regarding the display time of each image recorded in the buffer memory 103 in the process of S50 in FIG. Time intervals tb, td, and tbb Is the display time T of the target image P, the display time 前方 of the forward reference image Ρ, and the backward reference

cur cur rO, 0 rO, 0

Using the display time T of the image P, the display time 基準 of the reference image Ρ, and the display rl, 0 rl, 0 col col colref time T of the image。, the following formula is used.

colref

tb = T-T • (3)

rO, 0 cur

td = T-T (4)

rl, 0 rO, 0

tbb = T-T ••• (5)

col colref

Temporal direct prediction is performed for each of target area A and area B shown in Fig. 11.

cur cur

The two objects in the image from region A to region A, and region B

rO, 0 rl, 0 colref Prediction method assuming constant motion in the same direction in subregion B

[0020] It should be noted that in pmv (Predictor of Motion Vector) prediction, force vector direct decoding using a difference vector between an estimated motion vector and a prediction vector decodes a motion vector without using such a difference vector. Specifically, the forward prediction vector mvLO and the backward prediction vector mvLl are used as motion vectors in the target area A.

cur

[0021] The prediction image of the target area A is the forward reference image P indicated by the forward prediction vector mvLO.

region A on cur rO, 0 and region A on the backward reference image P pointed to by the backward prediction vector mvLl

Generated by performing motion compensation using rO, 0 rl, 0 rl, 0.

[0022] <Detailed configuration of prediction vector deriving unit 102>

The detailed configuration of the prediction vector deriving unit 102 for deriving a prediction vector using temporal direct prediction is shown in the block diagram of FIG.

The prediction vector derivation unit 102 includes a derivation method selection unit 201, a switch 202, a temporal direct prediction unit 203, a spatial direct prediction unit 204, a pmv prediction unit 205, and a zero vector output unit 206.

The derivation method selection unit 201 selects a prediction vector derivation method according to the prediction method input from the variable length code decoding unit 100 and the presence or absence of the reference motion vector recorded in the buffer memory 103. The switch 202 performs switching to the prediction vector derivation method selected by the derivation method selection unit 201. Temporal direct prediction unit 203 obtains a prediction vector by a method determined by temporal direct prediction, and outputs it to motion vector decoding unit 101. Space die The rect prediction unit 204 obtains a prediction vector by a method determined by spatial direct prediction, and outputs it to the motion vector decoding unit 101. When there is a difference vector, the pmv prediction unit 205 obtains a prediction vector by pmv prediction that encodes the difference vector between the estimated motion vector and the prediction vector, and outputs it to the motion vector decoding unit 101. Regardless of the input to the prediction vector deriving unit 102, the zero vector output unit 206 always outputs the zero vector to the motion vector decoding unit 101 as a prediction vector.

[0025] <Procedure for Deriving Predictive Betatones in Temporal Direct Prediction>

Next, the derivation procedure of the forward prediction vector mvLO and the backward prediction vector mvLl when the out-of-time prediction is used as the motion compensation prediction method will be described using the flowchart of the derivation procedure shown in FIG.

First, the derivation method selection unit 201 requests the reference motion vector mvCol from the buffer memory 103 (S21).

Next, the derivation method selection unit 201 determines the presence / absence of the reference motion vector mvCol in the reference region B on the buffer memory 103 based on the notification from the buffer memory 103 (S22).

col

[0028] When the reference motion vector mvCol is present, the process proceeds to the next process of S23, and when there is no reference motion vector mvCol, the process proceeds to the process of S25.

[0029] Next, the derivation method selection unit 201 switches the switch 202, and the temporal direct prediction unit 2

Select 03 (S23).

[0030] Next, the temporal direct prediction unit 203 obtains the reference motion beta acquired from the buffer memory 103.

mvCol is scaled with temporal information, that is, using equation (1) and equation (2), the forward prediction vector mvLO and the backward prediction vector mvLl are calculated, and the motion vector vector decoding unit 101 is output as the prediction vector. The process ends (S24).

[0031] When the notification that the reference motion vector mvCol is not in the buffer memory 103 is received in S22, the derivation method selection unit 201 determines that the reference region B force S is intra-encoded.

col

The switch 202 is switched and the zero vector output unit 206 is selected (S25).

Next, the zero vector output unit 206 sets both the forward prediction vector mvLO and the backward prediction vector mvLl as zero vectors, outputs the motion vector decoding unit 101 as a prediction vector, and ends the processing (S26). . [0033] In addition to the temporal direct prediction, the MPEG-4 AVC method can perform motion compensated prediction by spatial direct prediction and pmv prediction. However, the spatial direct prediction unit 204 and the pmv prediction unit that perform these predictions. Explanation of the operation 205 is omitted.

[0034] As described above, in the non-temporal out-of-time prediction method according to the prior art, the back reference image P is the base image P, and the base motion rl, 0 col col of the base region B on the base image P is used. If the col vector mvCol is available, a prediction vector close to the actual motion of the object in the video can be derived.

However, in the prediction vector derivation method according to the conventional technique, the backward reference image P

rl, 0 Reference region B force If the S intra code is set, the reference motion vector mvCol is zero.

col

It will be a vector. In this case, as shown in the conceptual diagram of FIG. 14, since the calculations of Equation (1) and Equation (2) are performed using the zero vector, the region pointed to by the forward prediction vector mvLO is the region A and pointed to by the backward prediction vector The region becomes region A, and the forward prediction vector r0, 0 rl, 0

Tol mvLO and backward prediction vector mvLl and the actual motion of target area A

cur

Often becomes large. Accordingly, there is a problem that the prediction efficiency is lowered.

Disclosure of the invention

[0036] The present invention has been made in view of the above-described problems, and the object thereof is not a zero vector even when the base region B force intra-encoding on the backward reference image P is performed. 0 col

The object is to realize a moving picture decoding apparatus and a moving picture encoding apparatus capable of deriving a forward prediction vector mvLO and a backward prediction vector mvLl to prevent a decrease in prediction efficiency.

[0037] In order to solve the above problem, a prediction vector deriving unit that derives a prediction vector of a processing target region on a processing target image using a reference motion vector of a reference image, and the processing using the prediction vector Motion vector decoding means for reconstructing a motion vector of the target area, and predicted image derivation means for deriving a predicted image of the processing target area from the already decoded image after decoding processing using the motion vector. The prediction vector deriving unit derives a prediction vector by temporal direct prediction, wherein the prediction vector deriving unit extracts at least two or more decoded images as the reference image. One of the candidate images based on a predetermined selection criterion that avoids the reference motion vector from becoming a zero vector. Reference image And a reference motion vector selection means for selecting, as a reference motion vector, a motion vector of a region spatially located at the same position as the processing target region on the reference image.

The processing target image refers to an image that is a target for reconstructing an image constituting a moving image from the state of encoded data.

[0039] The processing target area refers to a portion of the processing target image that is on the processing target image and is decoded in units of one decoding process.

[0040] Prediction beta is a vector used to derive a motion vector.

[0041] The motion vector is a vector used for deriving a prediction image when reconstructing an image constituting a moving image by a motion compensation prediction method.

[0042] A reference image is an image having a reference motion vector for obtaining a prediction vector.

[0043] The reference motion vector is a motion vector in the reference area of the reference image.

The reference area is an area on the reference image that is in the same spatial position as the processing target area.

[0045] The predicted image is an image from which the image is reconstructed from the prediction residual data included in the encoded data.

[0046] The already-decoded image is an image constituting a moving image derived by decoding and reconstructing encoded data.

[0047] Temporal direct prediction refers to scaling the reference motion vector based on the time interval between the display time of the processing target image, the display time of the reference image, and the display time of the decoded image indicated by the reference motion vector. And a prediction method for deriving a prediction vector.

[0048] The display time is time information included in the encoded data, and indicates at which time each decoded image constituting the moving image should be reproduced.

[0049] In the prior art, a prediction margin is derived using a decoded image whose display time is later than the processing target image as a reference image. However, when the decoded image is intra-coded, the prediction vector is zero. In this configuration, a reference image is selected from a plurality of candidate images.

[0050] According to the above configuration, the decoded image whose display time is later than the processing target image is the intra code. Even if the reference region is encoded and has no motion margin, the reference image can be selected from a plurality of already decoded images included in the candidate image, so there is little correlation with the motion of the object on the image. The zero vector increases the possibility of obtaining a prediction vector that reflects the motion of an object on the image, which has the effect of improving the prediction efficiency of the prediction image.

[0051] In order to solve the above problem, a prediction vector deriving unit for deriving a prediction vector of a processing target region on a processing target image using a reference motion vector of a reference image, and the processing target image and the encoding process From the already-encoded image, the motion vector estimating means for estimating the motion vector, the motion vector encoding means for encoding the motion vector using the prediction vector, and the already-encoded image, Prediction image deriving means for deriving a predicted image of the processing target region using the motion outer map, and the prediction vector deriving means is a moving image code for deriving the prediction outer map by temporal out-of-time prediction. In the encoding apparatus, the prediction margin deriving means sets at least two or more of the already-encoded images as candidate images of the reference image, and the reference motion vector becomes a zero vector. One of the candidate images is selected as the reference image based on a predetermined selection criterion, and a motion vector of a region located on the reference image at the same spatial position as the processing target region It is characterized by comprising reference motion vector selection means for selecting as a reference motion vector.

An already-encoded image is an image obtained by decoding and reconstructing an image from once encoded data among images constituting moving image data.

In the conventional technique, a prediction margin is derived using an already-encoded image whose display time is later than that of the processing target image as a reference image. When the already-encoded image is intra-encoded, a prediction vector Is a zero vector. In this configuration, a reference image is selected from a plurality of candidate images.

[0054] According to the above configuration, even when an already-encoded image whose display time is later than the processing target image is intra-encoded and the reference region does not have a motion margin, a plurality of existing images included in the candidate image are included. Since the reference image can be selected from the encoded image, it is highly possible to obtain a prediction vector that reflects the motion of the object on the image that is less than the zero vector that has little correlation with the motion of the object on the image. The effect of improving the prediction efficiency of the predicted image is achieved. To do.

Brief Description of Drawings

FIG. 1 is a flowchart showing a procedure by which a prediction vector deriving unit 112 derives a prediction vector.

FIG. 2 is a block diagram showing a configuration of a moving picture decoding apparatus 1 in the first embodiment.

FIG. 3 is a block diagram showing a detailed configuration of a prediction vector deriving unit 112.

[Fig.4] Prediction based on temporal direct prediction when the forward reference image P is the base image P

rO, 0 col

It is a conceptual diagram explaining the derivation | leading-out method of a non-measurement lure.

FIG. 5 is a conceptual diagram illustrating a method for deriving a prediction curve based on out-of-time prediction when a plurality of reference images are set as reference image candidates.

FIG. 6 is a conceptual diagram illustrating a prediction rule derivation method based on out-of-time prediction when a non-reference image is included in a standard image candidate.

[Fig. 7] Fig. 7 is a block diagram showing a configuration of a moving image encoding device 2 in the second embodiment.

FIG. 8 is a flowchart showing a processing procedure of the moving picture encoding apparatus 2 in the second embodiment.

FIG. 9 is a block diagram of a video decoding device 3 in the prior art.

FIG. 10 is a flowchart showing an outline of a decoding procedure of the moving picture decoding apparatus 3 in the prior art.

[Fig. 11] Prediction based on temporal direct prediction when the backward reference image P is the base image P.

rl, 0 col

It is a conceptual diagram explaining the derivation | leading-out method of an off-measurement rule.

FIG. 12 is a block diagram showing a detailed configuration of a prediction vector deriving unit 102 in the prior art.

FIG. 13 is a flowchart showing a procedure by which the prediction vector deriving unit 102 derives a prediction vector in the prior art.

FIG. 14 is a conceptual diagram showing a prediction margin when the reference region B force S intra prediction is performed in the prior art.

BEST MODE FOR CARRYING OUT THE INVENTION [0056] [First embodiment]

Hereinafter, the moving picture decoding apparatus according to the first embodiment of the present invention will be described.

[0057] <Configuration of video decoding device>

In the block diagrams of FIG. 2 and FIG. 3, the configuration of the moving picture decoding apparatus in the present embodiment is shown. The block diagram of FIG. 2 shows the overall configuration of the video decoding device 1, and the block diagram of FIG. 3 shows the detailed configuration of the prediction vector derivation unit (prediction vector derivation means) 112 shown in FIG. Is.

[0058] The elements shown in these figures are the same as the block diagrams (FIGS. 9 and 12) used in the description of the prior art except for those described below, and thus the description thereof is omitted.

In the block diagram shown in FIG. 2, the difference between the configuration of moving image decoding apparatus 1 of the present embodiment and the configuration of moving image decoding apparatus 3 according to the conventional technique is prediction vector deriving unit 112. Based on the prediction method input from the variable-length code decoding unit 100, the prediction vector deriving unit 112 derives a prediction vector using the decoded motion vector.

In the block diagram shown in FIG. 3, the difference between the configuration of the prediction vector derivation unit 112 of the present embodiment and the configuration of the prediction vector derivation unit 102 according to the prior art is a reference vector selection unit (reference motion vector selection). Means) 210 or not. The reference vector selection unit 210 selects an image to be the reference image P from a plurality of reference image candidates on the buffer memory 103, and

col

The reference motion vector mvCol corresponding to the selected reference image P is obtained from the derivation method selection unit 201 col

Notify

That is, the moving picture decoding apparatus 1 according to the present embodiment uses the reference picture P and the reference motion vector mvC col in a predetermined procedure from a plurality of reference picture candidates in the temporal direct prediction.

ol is determined, and a forward prediction vector mvLO and a backward prediction vector norm mvLl are derived.

A procedure for deriving the forward prediction vector mvLO and the backward prediction vector mvLl in the temporal direct prediction in the moving picture decoding apparatus 1 according to the present embodiment will be described with reference to the flowchart of FIG.

[0063] First, the derivation method selection unit 201 sends a reference motion vector mvC to the reference vector selection unit 210. Request ol (S l).

Next, after the reference vector selection unit 210 sets the back reference image P as the reference image P (P = P) in response to a request from the derivation method selection unit 201, the reference of the reference image P

rl, 0 col col rl, 0 col A motion vector mvCol is requested to the buffer memory 103 (S2).

Next, the derivation method selection unit 201 determines whether or not the reference motion vector mvCol exists in the buffer memory 103 based on the response from the buffer memory 103 (S3). In this case, the presence of the reference motion vector mvCol in the buffer memory 103 means that the reference motion vector mvCol exists in the reference region B of the backward reference image P.

rl, 0 col

[0066] When the reference motion vector mvCol exists in the buffer memory 103 (back reference image P

If rl is the reference image P (corresponding to the conceptual diagram in FIG. 11), the process proceeds to the next step S4.

, 0 col

If the reference motion vector mvCol does not exist in the buffer memory 103, the process proceeds to S6.

[0068] When it is determined in S3 or S7 that the reference motion vector mvCol is present in the buffer memory 103, the derivation method selection unit 201 switches the switch 202 and selects the time direct prediction unit 203 (S4 ).

[0069] Next, the temporal direct prediction unit 203 uses the reference motion vector mvCol acquired from the buffer memory 103 to calculate the forward prediction vector mvLO and the backward prediction vector mvLl according to Equation (1) and Equation (2). The calculation results are output as the forward prediction vector mvLO and the backward prediction vector mvLl to the motion vector decoding unit 101, and the process is terminated (S5).

[0070] If it is determined in S3 that the standard motion vector mvCol does not exist in the buffer memory 103, the standard vector selection unit 210 sets the forward reference image P as the standard image P.

After setting rO, 0 col (P = P), the reference motion vector mvCol of the reference image P is stored in the buffer memory 10 col rO, 0 col

Request to 3 (S6).

Next, the derivation method selection unit 201 determines whether or not the reference motion vector mvCol exists in the buffer memory 103 based on the response from the nonophor memory 103 (S7). In this case, the presence of the reference motion vector mvCol in the buffer memory 103 means that the reference motion vector mvCol exists in the reference region B of the forward reference image P.

rO, 0 col

[0072] When the reference motion vector mvCol exists in the buffer memory 103 (in the case of the conceptual diagram in FIG. 4) The process proceeds to S4. In the conceptual diagram of FIG. 4, the forward reference image P is used as the basis.

In the case of rO, 0 quasi-image P, the reference motion vector mvCol and the forward prediction vector mvLO and

丄

The relationship with the backward prediction vector mvLl is shown.

[0073] If the reference motion vector mvCol does not exist in the buffer memory 103, the process proceeds to the next S8.

[0074] When it is determined in S7 that the reference motion vector mvCol does not exist in the buffer memory 103, the derivation method selection unit 201 switches the switch 202 and selects the zero vector output unit 206 (S8).

Next, the zero vector output unit 206 force S and the zero vector are output to the motion vector decoding unit 101 as the forward prediction vector mvLO and the backward prediction vector mvLl, and the process ends. (S9)

According to the above procedure, when the back reference image P is set as the base image P, the base motion beta is

rl, 0 col

If mvCol is available, the prediction vector can be derived in the same manner as in the prior art. Also, if the base motion vector mvCol in the back reference image P is not available

rl, 0

For example, the standard motion vector mvCol should be used with the forward reference image P as the base image P.

rO, 0 col

Can do.

[0076] As in the case of using the backward reference image P as the base image P, the forward reference image P is

rl, 0 col rO, 0 When used as the reference image P, the image is also displayed in the target area A shown in the conceptual diagram of

The object col cur and the object in image in region B

cur rO, 0 It is assumed that uniform motion is performed in the same direction from region B to region B and from region B to region B. rl, 0 colrei rl, 0

Make a prediction.

[0077] Thus, in the conventional method, as shown in the conceptual diagram of Fig. 14, compared to a prediction method in which a prediction vector having no reference motion vector mvCol in the backward reference image P rl, 0 is a zero vector. In the prediction method according to the present invention, the prediction vector is derived using the standard motion vector mvCol of another reference image and used for image prediction.

cur

The movement at the time can be expressed accurately. Therefore, it is possible to make predictions with higher efficiency than the prediction methods based on the prior art.

[0078] <Reference image selection method when there are three or more reference image candidates> The present invention can also be applied when there are three or more reference image candidates.

[0079] For example, when there are three or more reference images, a plurality of reference images can be used as reference image candidates. As shown in the conceptual diagram of FIG. 5, three forward reference images P rO, 0

, Ρ, and Ρ, and one backward reference image Ρ, and these reference images are all rO, 1 rO, 2 rl, 0

The case where the reference image candidate is used will be described below.

[0080] When there are three or more reference image candidates, priorities among the candidates are determined using the time interval between the display time of each reference image candidate and the display time of the target image P.

cur

For example, rear reference images P are arranged in ascending order of time interval, and then forward reference images

rl, 0

P, P, and P are arranged in ascending order of time interval, and this order is the priority rO, 0 rO, 1 rO, 2

. According to this priority order, the priority order of each standard image candidate in FIG. 5 is the order of the backward reference image P, the forward reference image P, the forward reference image P, and the forward reference image P.

rl, 0 r0, 0 r0, 1 rO, 2

[0081] The reference margin selection unit 210 sets a reference image candidate having a high priority in the reference image P.

col Then, the presence or absence of the reference motion vector mvCol of the set reference image P is determined.

col

If the reference motion vector mvCol exists, the reference motion vector mvCol is selected. If it does not exist, the reference image candidate of the next priority is set as the reference image P, and the same processing is repeated according to the priority. After all the reference images P have been tried,

col

If the quasi-motion vector mvCol is not selected, the derivation method selection unit 201 is notified of that fact.

In the conceptual diagram shown in FIG. 5, the reference motion when the reference image P is set as the reference image P is shown.

rO, 1 col

The relationship between the forward vector mvCol, the forward prediction vector mvLO, and the backward prediction vector mvLl is shown. In this case as well, the object in the target area A and the area B

Assuming that an object as a cur cur image is moving in the same direction from region A to region A and from region B to region B rO, 0 rl, 0 colref r in the same direction, respectively (1) And the formula (2)

Ten

Torr can be calculated.

[0083] When there are three or more reference image candidates, any one of the reference image candidates is set as the reference image P col, and the reference motion vector mvCol corresponding to the reference image P is selected.

丄

As shown in the conceptual diagram, it is more likely that a prediction vector having a high correlation with the actual motion of the object in the image can be derived compared to the case where the reference motion vector mvCol is set to zero vector nore. Become. Therefore, more efficient prediction is possible.

[0084] <Prediction method for including non-reference image in standard image candidate>

In addition to the reference image, a non-reference image that is not used for derivation of a predicted image in the related art may be included in the reference image candidate. In this case, all already decoded images can be candidates for the reference image. As shown in the conceptual diagram of FIG. 6, the forward reference image P and the backward reference image P rO, 0

Rl, 0 nr when the non-reference image Ρ is included in the standard image candidates in addition to the two reference images

And explain.

[0085] Even when the reference image candidate includes the non-reference image P, the reference vector selection unit 210

nr

The operation is the same as that already described. In other words, the reference motion col is determined by determining the presence or absence of the reference motion vector mvCol for each image set as the reference image P in order according to the priority order.

Select the vector mvCol.

[0086] As a method of determining the priority when the reference image candidate includes the non-reference image P, for example,

nr

Regardless of whether it is a reference image or non-reference image, the display time of the target image P is displayed.

cur

There is a method of assigning priorities to images in the order of time. Also, images that are reference images are arranged in order of display time close to the display time of the target image P, and then are non-reference images.

cur

Arrange the images in order of display time close to the display time of the target image P, and set the priority

cur

There is also a law.

[0087] In the conceptual diagram of FIG. 6, when the non-reference image P is set to the standard image P (P = P).

The relationship between the reference motion vector mvCol, the forward prediction vector mvLO, and the backward prediction vector mv L1 in nr col col nr is shown. In this case as well, the object and area that are in the target area A as an image

cur

Assuming that an object as an image in region B is moving in the same direction in the same direction from region A to region A and from region B cur rO, 0 rl, 0 coir to reference region B, respectively. (1) and the formula (ef rl, 0

2) Calculate the predicted margin.

[0088] By including the non-reference image in the standard image candidate, there are more standard image candidates whose display time is close to the display time of the target image P than when only the reference image is included in the standard image candidate.

cur

Become. If the interval between the display time of the target image P and the display time of the reference image P is narrow,

cur col

Since the distance between the target area A and the area B on the target image P becomes shorter, these 2 cur cur cur

It is easy to assume that two regions are moving at the same speed in the same direction. Therefore, Since there is a high possibility that a motion vector having a high correlation with the actual motion of an object in the image can be derived, more efficient prediction is possible than when only the reference image is used as a reference image candidate.

[0089] When a non-reference image is used as a standard image candidate, at least a motion vector corresponding to the candidate non-reference image and display time information of the non-reference image are recorded in the buffer memory 103. Need to be.

[0090] <Method using only 参照 reference image for prediction image derivation>

In the above description, the prediction image of the target area A is used in the temporal direct prediction.

cur

Two images, a forward reference image P and a backward reference image P, are used for image derivation.

rO, 0 rl, 0

Only one of the reference images may be used.

For example, as shown in the conceptual diagram of FIG. 4, the forward reference image P is predicted as the standard image P.

r0, 0 col

When deriving a vector, a prediction image of the target area A is generated from only the area A.

r0, 0 cur

Also good. In this case, the object in the reference area B as an image is from area B to area B.

col colref cur Direction of possibility of constant velocity movement Region B with longer time interval Constant force to region B

colref rl, 0

Higher than the possibility of exercising. In other words, the forward prediction vector mvLO is more likely to be more accurate than the backward prediction vector mvLl. Therefore, only region A was used

rO, 0

In some cases, the prediction efficiency is improved.

[0092] Further, the backward reference image P is set as the base image P, and the base motion beta is shown.

rl, 0 col

If nore mvCol refers to an image whose display time is later than the back reference image P,

rl, 0

As in the case of deriving the prediction vector from only the reference image P, only the region A is predicted.

rO, 0 rl, 0 Prediction efficiency is improved by using it for derivation of measured images.

[0093] [Second Embodiment]

Hereinafter, the moving picture coding apparatus according to the second embodiment of the present invention will be described. <The configuration of the video encoding device>

In the block diagram of FIG. 7, the configuration of moving picture coding apparatus 2 according to the present embodiment is shown. The block diagram of FIG. 7 shows the overall configuration of the video encoding device 2.

[0094] The moving image encoding apparatus 2 includes an image encoding unit 121, a predicted image derivation unit 105, a motion vector estimation unit (motion vector nore estimation means) 122, a nother memory 103, an image decoding unit 104, A tuttle deriving unit 112, a prediction scheme control unit 124, a variable length coding unit 120, and a motion vector coding unit (motion vector coding means) 123 are configured.

The elements shown in this figure are the same as those shown in the block diagram (FIG. 2) of the moving image decoding apparatus 1 in the first embodiment, except for those described below, and thus the description thereof is omitted.

The variable length coding unit 120 includes prediction residual data input from the image coding unit 121, a difference vector input from the motion vector coding unit 123, and a prediction method input from the prediction method control unit 124. Encoding information is variable-length encoded and output to the outside. The image encoding unit 121 obtains prediction residual data using the moving image data input from the outside and the predicted image input from the predicted image deriving unit 105, and the image decoding unit 104 and the variable-length code encoding unit 120. Output to. The motion vector estimation unit 122 estimates the motion vector nore using the externally input moving image data and the reference image in the buffer memory 103, and uses the obtained motion vector nore as the predicted image deriving unit 105, the buffer memory 103, the prediction The vector deriving unit 112, the prediction scheme control unit 124, and the motion vector encoding unit 123 are used. The motion vector encoding unit 123 uses a motion vector input from the motion vector estimation unit i 22, a prediction vector input from the prediction vector derivation unit 112, and a prediction vector input from the prediction method control unit, using a difference vector. Is output to the variable length encoding unit 120. The prediction method control unit 124 sets a prediction method for deriving a prediction vector based on the motion vector input from the motion vector estimation unit 122, and sets the prediction method as a prediction vector deriving unit 112, a variable length coding unit 120, and It outputs to the motion vector code part 123.

The outline of the encoding processing procedure in the moving image encoding apparatus 2 will be described below with reference to the flowchart shown in FIG.

[0098] First, the motion vector estimation unit 122 performs motion estimation using the input moving image data and the reference image in the buffer memory 103, and motion vector for each region of the image to be encoded. Ask for. The motion vector estimation unit 122 records the obtained motion vector in the buffer memory 103, and outputs the motion vector to the prediction vector deriving unit 112, the prediction scheme control unit 124, and the motion vector coding unit 123 (S100).

Next, the prediction scheme control unit 124 receives the motion vector input from the motion vector estimation unit 122. Are used to determine the prediction method as one of temporal direct prediction, spatial direct prediction, or pmv prediction, and the prediction vector deriving unit 112, variable-length encoding unit 120, and motion vector encoding unit It outputs to 123 (S110).

Next, based on the motion vector input from the motion vector estimation unit 122 and the prediction method input from the prediction method control unit 124, the prediction vector deriving unit 112 performs the first embodiment. In accordance with the prediction rule derivation procedure described in the above-mentioned state, the prediction error is calculated and output to the buffer memory 103 and the motion vector code key unit 123 (S120).

Next, the motion vector encoding unit 123 obtains a difference vector. Prediction method control unit 124 force For input prediction method power Spmv prediction, based on the motion vector input from the motion vector estimation unit 122 and the prediction vector input from the prediction vector derivation unit 112, the difference Calculate the vector. Prediction method input from prediction method control unit 124 In the case of temporal direct prediction or spatial direct prediction, the difference vector is set to zero vector. The motion vector encoding unit 123 outputs the obtained difference vector to the variable length encoding unit 120 (S130).

Next, the predicted image deriving unit 105 performs motion compensation using the motion vector input from the motion vector estimating unit 122 and the reference image in the buffer memory 103 to obtain a predicted image. The image is output to the image encoding unit 121 and the image decoding unit 104 (S140).

Next, the image encoding unit 121 calculates prediction residual data based on the moving image data input from the outside and the predicted image input from the predicted image deriving unit 105, and the image decoding unit 10 4 And it outputs to the variable length encoding part 120 (S150).

[0104] Next, the image decoding unit 104 uses the prediction residual data input from the image encoding unit 121 and the prediction image input from the prediction image deriving unit 105 as an encoded image. To reconfigure (S 160). The reconstructed image is recorded in the buffer memory 103. The already-encoded image is used as a reference image when the predicted vector is derived, and is used as a reference image when the predicted image is derived.

[0105] Next, the variable length coding unit 120 inputs the prediction residual data input from the image coding unit 121, the difference vector input from the motion vector coding unit 123, and the prediction scheme control unit 124. A variable-length code and a video code as a code data. Output to the outside of the device 2 (S170).

[0106] <Supplementary Information on Second Embodiment>

As described above, the moving picture coding apparatus 2 according to the present embodiment derives a prediction vector using a plurality of reference image candidates according to the prediction margin derivation procedure described in the first embodiment. . Since the moving picture is encoded using the derived prediction vector, the moving picture can be encoded with high efficiency.

[0107] In the process of S120 (derivation of the prediction vector) in the flowchart shown in FIG.

As described in the first embodiment, the prediction efficiency is improved by deriving a prediction vector using three or more reference image candidates.

Similarly, in the process of S120, as described in the first embodiment, the prediction efficiency is improved by deriving the prediction margin by including the non-reference image in the standard image candidate.

[0109] The prediction vector derivation method in the temporal direct prediction described above can also be used for derivation of a prediction vector when pmv prediction is used as a prediction method.

[0110] <Supplementary information>

The present invention is not limited to the above-described embodiments, and various modifications can be made within the scope shown in the claims, and embodiments obtained by appropriately combining technical means disclosed in different embodiments. Is also included in the technical scope of the present invention.

[0111] The candidate image has a display time later than that of the processing target image and a display time that is earlier than the processing target image and has a display time earlier than that of the processing target image and the processing target image. Is the nearest decoded image, and the predetermined selection criterion is a motion vector of an area located spatially at the same position as the processing target area on the candidate image whose display time is later than the processing target image. If the candidate image is selected as the reference image, and the motion vector does not exist, the candidate image is displayed on the candidate image earlier in display time than the processing target image. In the case where there are motion vectors of regions located at the same spatial position, it is preferable that the criterion is that the candidate image is selected as the reference image. Accordingly, in the configuration, the reference motion vector selection unit first determines whether or not the already-decoded image immediately after the processing target image can be selected as the reference image. If it can be selected, the predicted image is derived in the same way as in the conventional technology. If it cannot be selected, it is next determined whether or not the already decoded image immediately before the processing target image can be selected as the reference image. If it can be selected, a predicted image is derived using the decoded image as a reference image.

[0113] According to the above configuration, since the decoded image having the closest display time to the processing target image is selected as the reference image, the time interval between the processing target image and the reference image is short. Therefore, the object on the image, which is the premise of time-directed prediction, has a higher probability that the assumption that it is moving at the same speed in the same direction will be increased, so that it is possible to improve the prediction efficiency of the predicted image.

[0114] In addition, the predetermined selection criterion includes priorities of decoded images that are included in the candidate images and that have a display time later than the processing target image, in order of display time closer to the display time of the processing target image. The decoded images included in the candidate images, the display times of which are earlier than the processing target images, are given priorities in the order of the display times closer to the display times of the processing target images. In the order of the priorities, the determination of the presence or absence of a motion vector in an area spatially located at the same position as the processing target area on the candidate image is repeated until the motion vector is present. In such a case, it is preferable that the criterion is that the candidate image is selected as the reference image.

[0115] In this configuration, the reference motion vector selection unit first assigns priorities to a plurality of already-decoded images, and determines whether or not the reference image can be selected in descending order of priority. . This priority order is a priority order in which the decoded image having a high possibility that the prediction vector reflects the actual motion of the object is prioritized.

[0116] Thus, according to the configuration described above, it is possible to select any of the candidate images as a reference image and select a reference motion vector corresponding to the reference image, so that the reference motion vector is zero vector. As a result, there is a high possibility that a prediction vector having a high correlation with the actual motion of the object in the image can be derived, so that more efficient prediction can be achieved.

[0117] Further, the reference motion vector selection means has a display time earlier than that of the processing target image. When a previously decoded image is selected as the reference image, the predicted image deriving unit preferably derives a predicted image using only the forward reference image whose display time is closest to the processing target image.

[0118] A reference image is a decoded image having an area pointed to by a motion vector.

[0119] The forward reference image refers to a decoded image having a display time earlier than the processing target image among the decoded images having the area indicated by the motion vector.

[0120] The backward reference image is a decoded image having a display time later than that of the processing target image, among the decoded images having the area indicated by the motion vector.

[0121] The forward prediction vector is a prediction vector indicating the forward reference image from the processing target region.

[0122] The backward prediction vector is a prediction vector indicating the backward reference image from the processing target region.

[0123] In this configuration, when viewed from the display time, there is a forward reference screen pointed to by the base motion vector first, then a base image, next a processing target image, and then a back reference image. ,become.

Thus, according to the above configuration, when a prediction vector is derived using a forward reference image as a base image, a prediction image of the processing target region is generated only from the region on the base image pointed to by the forward prediction vector. In this case, the possibility that an object as an image in the reference area moves at a constant speed from the area on the forward reference image pointed to by the reference motion vector to the area on the processing target image points to the reference motion vector having a longer time interval. Compared to the possibility of moving at a constant speed from the area on the forward reference image to the area on the backward reference image.

[0125] Further, since the prediction vector is obtained using the time interval between the forward reference image and the processing target image that is shorter than the time interval between the forward reference image and the backward reference image, the image that is an assumption of temporal direct prediction. The possibility that the condition that the upper object moves at the same speed in the same direction increases. Therefore, there is an effect that the prediction efficiency of the predicted image can be improved.

[0126] Further, the decoded image is preferably a reference image and a non-reference image.

[0127] A non-reference image is a decoded image that is not used for deriving a predicted image.

[0128] In this configuration, by including a non-reference image in the candidate image, only the reference image is a candidate. Compared to the case where the image is included in the image, the number of already decoded images whose display time is close to the display time of the processing target image increases.

Thus, according to the above configuration, when the interval between the display time of the processing target image and the display time of the reference image is narrow, the area on the reference image pointed to by the target area and the reference movement belt Since the distance on the image to be processed becomes shorter, it is easy to assume that these two areas are moving at the same speed in the same direction. Therefore, there is a high possibility that a motion vector having a high correlation with the actual movement of the object in the image can be derived, so that the prediction efficiency can be further improved compared to the case where only the reference image is set as the reference image candidate. Play.

[0130] Further, the candidate image has a display time later than the processing target image and a display time earlier than the processing target image and an already encoded image whose display time is closest to the processing target image and the processing target image. The already-encoded image having the closest display time, and the predetermined selection criterion is an area of the candidate image that is later in display time than the processing target image and is spatially located at the same position as the processing target region. When the motion vector exists, the candidate image is selected as the reference image, and when the motion vector does not exist, the display time is earlier than the processing target image, and the processing target region is displayed on the candidate image. If there is a motion vector in a region located in the same spatial position, it is preferable that the criterion is that the candidate image is selected as the reference image.

In this configuration, the reference motion vector selection unit first determines whether or not the already-encoded image immediately after the processing target image can be selected as the reference image. If it can be selected, the prediction image is derived as in the conventional technique. If it cannot be selected, it is next determined whether or not the already-encoded image immediately before the processing target image can be selected as the reference image. If it can be selected, a predicted image is derived using the already-encoded image as a reference image.

Thus, according to the above configuration, the pre-signed image having the closest display time to the processing target image is selected as the reference image, so the time interval between the processing target image and the reference image is short. Therefore, the object on the image, which is the premise of temporal direct prediction, increases the probability that the assumption that the object will move at the same speed in the same direction will be satisfied, so that it is possible to improve the prediction efficiency of the predicted image.

[0133] Further, the predetermined selection criterion is based on the processing target image included in the candidate image. Prior coded images with a later display time are prioritized in the order in which the display time is closer to the display time of the processing target image, and are displayed from the processing target image included in the candidate image following the priority order. Priorities are assigned to encoded images with earlier times in the order in which the display time is closer to the display time of the processing target image, and in the order of the priority, spatially with the processing target area on the candidate image. The determination of the presence / absence of the motion vector of the region located at the same position is repeated until the motion vector is present, and if the motion vector is present, the candidate image may be selected as the reference image. preferable.

[0134] In this configuration, the reference motion vector selection means first assigns priorities to a plurality of already-encoded images, and determines whether or not the reference images can be selected in descending order of priority. . This priority order is a priority order for a pre-signed image with a high possibility that the prediction vector reflects the actual motion of the object.

Thus, according to the above configuration, as compared with the case where the reference motion vector is set to zero vector nore by selecting any power of the candidate image as the reference image and selecting the reference motion vector corresponding to the reference image. As a result, there is a high possibility that a prediction vector having a high correlation with the actual motion of the object in the image can be derived, so that more efficient prediction can be achieved.

[0136] Further, when the reference motion vector selection unit selects an already-encoded image having a display time earlier than the processing target image as the reference image, the predicted image deriving unit displays the processing target image on the processing target image. It is preferable that the predicted image is derived using only the forward reference image at the latest time.

[0137] In this configuration, when viewed from the display time, there is a forward reference screen pointed to by the base motion vector first, then a base image, next a processing target image, and then a back reference image. ,become.

Thus, according to the above configuration, when the prediction vector is derived using the forward reference image as the reference image, the prediction image of the processing target region is generated only from the region on the reference image pointed to by the forward prediction vector. In this case, the possibility that an object as an image in the reference region moves at a constant speed from the region on the forward reference image indicated by the reference motion vector to the region on the processing target image has a longer time interval. The area on the forward reference image pointed to by the reference motion vector Force Higher than the possibility of moving at a constant speed to the region on the rear reference image.

[0139] Furthermore, since the prediction vector is obtained using the time interval between the forward reference image and the processing target image, which is shorter than the time interval between the forward reference image and the backward reference image, this is an assumption of temporal direct prediction. Therefore, the possibility that the condition that the object on the image moves at the same speed in the same direction increases. Therefore, there is an effect that the prediction efficiency of the predicted image can be improved.

[0140] Further, it is preferable that the already-encoded image is a reference image and a non-reference image.

[0141] In this configuration, by including the non-reference image in the candidate image, the number of already-encoded images whose display time is close to the display time of the processing target image is larger than when only the reference image is included in the candidate image.

[0142] Thus, according to the above configuration, if the interval between the display time of the processing target image and the display time of the reference image is narrow, the region on the reference image pointed to by the target region and the reference motion belt Since the distance on the image to be processed becomes shorter, the assumption that these two regions are moving at the same speed in the same direction is likely to hold. Therefore, there is a high possibility that a motion vector having a high correlation with the actual movement of the object in the image can be derived, so that the prediction efficiency can be further improved compared to the case where only the reference image is set as the reference image candidate. Play.

Note that the moving picture decoding apparatus according to the present invention includes a prediction vector deriving unit that derives a motion vector of a processing target region on a processing target image using a motion vector of a decoded image, and the prediction vector. And a predicted image deriving unit for deriving a predicted image of the processing target region from the decoded image using the prediction image deriving unit, wherein the prediction margin deriving unit uses at least two or more decoded images as a reference image One of the candidates is selected as a reference image based on a predetermined selection criterion, and a motion vector of a region spatially located at the same position as the processing target region on the reference image is selected as a reference motion vector. Reference motion vector selection means for selecting as a token is provided, and the prediction motion vector is derived by scaling the reference motion vector based on the display time interval between the images. It ’s good enough

[0144] In addition to the above-described configuration, the moving image decoding apparatus according to the present invention is configured so that the reference motion vector selecting means has a display time later than the processing target image and closest to the processing target image. Display time is faster than the decoded image and the processing target image, and the processing target image The already decoded image with the closest display time is set as the reference image candidate, and if there is a motion vector of an area located at the same position as the processing target area on the reference image candidate with a display time later than the processing target image, the motion When the outer vector is selected as the reference motion vector and the motion vector does not exist, there is a motion vector in the region located at the same position as the processing target region on the reference image candidate whose display time is earlier than the processing target image. In such a case, the motion vector may be selected as a reference motion vector.

[0145] Further, in the moving picture decoding apparatus according to the present invention, in addition to the above-described configuration, the reference motion vector selection means sets a plurality of decoded images as reference image candidates and is included in the reference image candidates. Arrange the decoded images whose display time is later than that of the processing target image in order of display time closer to the processing target images, and then display the decoded image earlier in the processing target image than the processing target images included in the reference image candidate. Are arranged in the order of closeness to be the priority order of the reference image candidates, the reference image candidates are set as the reference images in the order of the priority order, and there is a motion vector of a region located on the reference image at the same position as the processing target region. In such a case, the motion vector may be selected as a reference motion vector.

[0146] Further, in addition to the above configuration, the moving picture decoding apparatus according to the present invention uses the reference motion vector selection means as a reference motion vector, with a decoded image having a display time later than that of the processing target image as a reference image. Is selected, the prediction image is derived using the forward reference image whose display time is closest to the processing target image and the backward reference image whose display time is closest to the processing target image, and the display time is shorter than the processing target image. In the case where the reference motion beta is selected using the image with the fastest time as the reference image, the prediction image may be derived using only the forward reference image whose display time is closest to the processing target image.

[0147] Furthermore, in addition to the above configuration, the moving image decoding apparatus according to the present invention may be configured such that the reference motion vector selection means uses a reference image as the already decoded image. Yo!

[0148] In addition to the above configuration, the moving image decoding apparatus according to the present invention is characterized in that the reference motion vector selection means uses a reference image and a non-reference image as the already-decoded image. But you can.

[0149] Furthermore, the moving image encoding apparatus according to the present invention is a method for encoding an already encoded image that has been subjected to the encoding process. Prediction vector deriving means for deriving a motion vector of a processing target region on the processing target image using motion vector nore, and prediction image derivation for deriving a prediction image of the processing target region from an already encoded image using the prediction vector. The prediction vector derivation means uses at least two or more already-encoded images as candidates for the reference image, and selects from among the candidates based on a predetermined selection criterion: A reference motion vector selecting means for selecting a motion vector of a region that is spatially located at the same position as the processing target region on the reference image as a reference motion vector. The prediction vector may be derived by scaling the display based on the display time interval between the images.

[0150] In addition to the above configuration, the moving image encoding apparatus according to the present invention is such that the reference motion vector selection means has a display time later than the processing target image and the display time closest to the processing target image. An already-encoded image whose display time is earlier than that of the already-encoded image and the processing target image and whose display time is closest to the processing target image is set as the reference image candidate, and is processed on the reference image candidate whose display time is later than that of the processing target image. If there is a motion vector of a region located at the same position as the target region, the motion vector is selected as a reference motion vector, and if there is no motion vector, the display time from the processing target image is selected. If there is a motion vector of a region located at the same position as the processing target region on an early reference image candidate, the motion vector may be selected as the reference motion vector.

[0151] Further, in the moving image encoding device according to the present invention, in addition to the above-described configuration, the reference motion vector selection means sets a plurality of already-encoded images as reference image candidates and includes them in the reference image candidates. Arrange the pre-encoded images whose display time is later than the processing target image in order of display time close to the processing target images, and then select the pre-encoded images whose display time is earlier than the processing target images included in the reference image candidates. Are arranged in descending order of display time and set as the priority order of the reference image candidates, the reference image candidates are set as reference images in the order of the priority order, and the motion vector of the region located at the same position as the processing target region on the reference image If the motion vector exists, the motion vector may be selected as the reference motion vector.

[0152] In addition to the above configuration, the moving image encoding apparatus according to the present invention includes the reference motion vector. When the reference motion vector is selected with the pre-encoded image whose display time is slower than the processing target image as the reference image, the forward reference image and the processing target image whose display time is closest to the processing target image. When a reference image is derived using the back reference image with the closest display time for the image, and the reference motion beta is selected with the image having a display time earlier than the processing target image as the reference image, the display time for the processing target image is displayed. The configuration may be such that the predicted image is derived using only the forward reference image that is closest.

[0153] Further, the moving picture encoding apparatus according to the present invention may be configured such that, in addition to the above-described configuration, the reference motion vector selection means uses a reference image as the already-encoded image.

[0154] Further, in addition to the above configuration, the moving picture encoding apparatus according to the present invention is characterized in that the reference motion vector selection means uses a reference image and a non-reference image as the already-encoded image. The structure to do may be sufficient.

[0155] Finally, each block of the moving picture decoding device 1 and the moving picture coding device 2, particularly the derivation method selection unit 201, the reference vector selection unit 210, the temporal direct prediction unit 203, the spatial direct prediction unit 204, and pmv prediction And the zero vector output unit 206 may be configured by hardware logic, or may be realized by software using a CPU as follows.

That is, the video decoding device 1 and the video encoding device 2 are a CPU (central processing unit) that executes instructions of a control program that realizes each function, and a ROM (read only memory) that stores the program. In addition, a random access memory (RAM) for expanding the program, a storage device (recording medium) such as a memory for storing the program and various data, and the like are provided. The object of the present invention is to provide program codes (execution format program, intermediate code program, source program) for control programs of the video decoding device 1 and the video encoding device 2 which are software that realizes the functions described above. Is supplied to the moving image decoding device 1 and the moving image encoding device 2 and the computer (or CPU or MPU) is recorded on the recording medium. It can also be achieved by reading and executing the code.

[0157] Examples of the recording medium include a tape system such as a magnetic tape and a cassette tape, and a floppy disk. Disk system including magnetic disk such as P (registered trademark) disk / hard disk and optical disk such as CD-ROM / MO / MD / DVD / CD-R, card system such as IC card (including memory card) / optical card Alternatively, a semiconductor memory system such as mask ROM / EPROM / EEPROM / flash ROM can be used.

[0158] Further, the moving picture decoding apparatus 1 and the moving picture encoding apparatus 2 may be configured to be connectable to a communication network, and the program code may be supplied via the communication network. The communication network is not particularly limited. For example, the Internet, intranet, extranet, LAN, ISDN, VAN, CATV communication network, virtual private network, telephone line network, mobile communication network, A satellite communication network or the like can be used. Also, the transmission medium constituting the communication network is not particularly limited. For example, even in the case of wired communication such as IEEE1394, USB, power line carrier, cable TV line, telephone line, ADSL line, etc., infrared rays such as IrDA remote control. Bluetooth (registered trademark), 802.11 wireless, HDR, mobile phone network, satellite line, terrestrial digital network, etc. can also be used. The present invention can also be realized in the form of a computer data signal embedded in a carrier wave, in which the program code is embodied by electronic transmission.

[0159] As described above, in the video decoding device according to the present invention, the prediction vector deriving unit uses at least two or more of the already-decoded images as candidate images of the reference image, and provides a reference motion vector. Is selected as the reference image based on a predetermined selection criterion that avoids becoming a zero vector, and is positioned spatially at the same position as the processing target region on the reference image. And a reference motion vector selection means for selecting a motion vector of a region to be used as a reference motion vector signal.

[0160] Therefore, even when the decoded image power S-intra-coded later in the display time than the processing target image is encoded and the reference region has no motion margin, a plurality of decoded images included in the candidate image are not included. Since a reference image can be selected from among them, there is little correlation with the motion of the object on the image.There is a high possibility of obtaining a prediction vector that reflects the motion of the object on the image beyond the zero vector. There is an effect that the efficiency can be improved.

[0161] In addition, as described above, in the moving picture encoding apparatus according to the present invention, the prediction vector deriving means uses at least two or more already-encoded images as the reference image candidate images, and One of the candidate images is selected as the reference image based on a predetermined selection criterion that prevents the quasi-motion betaton from becoming a zero vector, and the processing target region and space are selected on the reference image. And a reference motion vector selection means for selecting a motion vector of a region located at the same position as a reference motion vector.

[0162] Therefore, even when an already-encoded image whose display time is later than that of the processing target image is intra-encoded and the reference region does not have a motion vector, a plurality of already-encoded images included in the candidate image are included. Since the reference image can be selected, it is highly possible to obtain a prediction vector that reflects the motion of the object on the image that is less than zero vector, which has little correlation with the motion of the object on the image. There is an effect that the efficiency can be improved. Industrial applicability

[0163] By using the moving picture decoding apparatus 1 and the moving picture encoding apparatus 2 according to the present invention, it is possible to improve the prediction efficiency of the prediction vector, that is, an apparatus for encoding or decoding a moving picture, that is, It can be suitably used for mobile terminal devices, mobile phones, television receivers, multimedia devices, and the like.

Claims

The scope of the claims

[1] A prediction margin deriving means for deriving a prediction vector of a processing target region on the processing target image using a reference motion vector of the reference image;

Motion vector decoding means for reconstructing a motion vector of the processing target region using the prediction vector;

Prediction image deriving means for deriving a prediction image of the processing target area from the already decoded image after decoding processing using the motion vector;

With

The prediction vector deriving unit is a video decoding device that derives a prediction vector by temporal direct prediction,

The prediction vector deriving means includes:

At least two or more of the already decoded images are used as candidate images for the reference image, and one of the candidate images is selected based on a predetermined selection criterion that avoids the reference motion vector becoming a zero vector. Select as reference image,

It is characterized by comprising reference movement margin selection means for selecting a movement margin of an area located spatially at the same position as the processing target area on the reference image as a reference movement margin. Video decoding device.

[2] The candidate image is

A decoded image having a display time later than that of the processing target image and closest to the processing target image.

and

A decoded image whose display time is earlier than the processing target image and whose display time is closest to the processing target image;

The predetermined selection criterion is:

If there is a motion vector of an area located at the same spatial position as the processing target area on the candidate image whose display time is later than the processing target image, the candidate image is selected as the reference image,

When the motion vector does not exist, On the candidate image whose display time is earlier than the processing target image, if there is a motion vector of a region located at the same spatial position as the processing target region, selecting the candidate image as the reference image, 2. The moving picture decoding apparatus according to claim 1, wherein the moving picture decoding apparatus is a standard.

[3] The predetermined selection criterion is:

Prioritize decoded images that are included in the candidate images and that have a display time later than the processing target image, in order from the display time closer to the display time of the processing target image.

Priorities of decoded images that are included in the candidate images and that are earlier in display time than the processing target image are given priority in order of display time close to the display time of the processing target image, and in the order of priority, The determination of the presence / absence of a motion vector in a region spatially located at the same position as the processing target region is repeated until the motion vector exists,

2. The moving picture decoding apparatus according to claim 1, wherein when there is the motion vector, the candidate image is selected as the reference image.

[4] The reference motion vector selection means includes:

When a decoded image having a display time earlier than the processing target image is selected as the reference image,

The predicted image derivation means includes

4. The moving picture decoding apparatus according to claim 1, wherein a predicted image is derived using only a forward reference image whose display time is closest to the processing target image. 5.

[5] The moving picture decoding apparatus according to any one of claims 1 to 4, wherein the already decoded pictures are a reference picture and a non-reference picture.

[6] A prediction margin deriving means for deriving a prediction vector of a processing target region on the processing target image using a reference motion vector of the reference image;

Motion vector estimation means for estimating a motion vector using the processing target image and the already-encoded image that has been subjected to the encoding process; A motion vector encoding means for encoding the motion vector using the prediction vector;

Prediction image deriving means for deriving a prediction image of the processing target region from the already-encoded image using the motion vector;

With

The prediction margin deriving means is a video encoding device for deriving a prediction margin by temporal out-of-time prediction,

The prediction vector deriving means includes:

Among the candidate images, based on a predetermined selection criterion, wherein at least two or more of the already-encoded images are candidate images of the reference image and the reference motion vector is prevented from becoming a zero vector: Select the images as the reference image,

A moving picture coding apparatus comprising reference motion vector selection means for selecting, as a reference motion vector, a motion vector in a region spatially located at the same position as the processing target region on the reference image .

The candidate image is

An already-encoded image whose display time is later than the processing target image and whose display time is closest to the processing target image

and

An already-signed image whose display time is earlier than the processing target image and whose display time is closest to the processing target image;

The predetermined selection criterion is:

If there is a motion margin of an area located at the same spatial position as the processing target area on the candidate image whose display time is later than the processing target image, the candidate image is selected as the reference image. ,

When the motion vector does not exist,

When there is a motion margin of an area located at the same spatial position as the processing target area on the candidate image whose display time is earlier than the processing target image, the candidate image is selected as the reference image. The standard according to claim 6, wherein Video encoding device.

[8] The predetermined selection criterion is:

The encoded image included in the candidate image, the display time of which is later than the processing target image, the display time is closer to the display time of the processing target image, prioritized in order, followed by the priority order,

Pre-coded images included in the candidate images, the display times of which are earlier in display time than the processing target images are given priority in order, the display times being closer to the display times of the processing target images, The determination of the presence or absence of a motion vector in an area located spatially at the same position as the processing target area on the candidate image is repeated until the motion vector exists.

7. The moving picture encoding apparatus according to claim 6, wherein when there is the motion vector, the candidate image is selected as the reference image.

[9] The reference motion vector selection means includes:

When an already-encoded image whose display time is earlier than the processing target image is selected as the reference image,

The predicted image derivation means includes

9. The moving picture encoding apparatus according to claim 6, wherein a predicted image is derived using only a forward reference image whose display time is closest to the processing target image.

10. The moving picture coding apparatus according to any one of claims 6 to 9, wherein the already coded pictures are a reference picture and a non-reference picture.