JP2003219417A - Method and device for encoding moving image, program for executing the method, and recording medium recording the program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To reduce memory band widths through moving image encoding in which reference padding is performed. <P>SOLUTION: A moving image processing system performs reference padding in which pixel values are substituted outside a VOP (a shape existing area of pixel signals), inputs pixel signals and shape signals, and outputs the inside of the VOP by encoding the inside from the shape signals. In this case, a third means 22 classifies the shape signals into blocks at the boundary of the VOP (state 1), blocks on the outside of the VOP in which one of adjacent blocks in four directions of upward, downward, leftward, and rightward directions has pixels in the VOP (state 2), or blocks on the outside of the VOP other than the above-mentioned blocks (step 3). A fourth means 24 decides the padding method of reference images as VOP boundary padding in which padding in the horizontal and vertical direction is performed in the state 1, expansion padding in which padding is performed by using the pixels in one of adjacent blocks in four directions of upward, downward, leftward, and rightward directions in the state 2, or fixed-value padding in the state 3 based on the state from the third means 22. <P>COPYRIGHT: (C)2003,JPO

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention receives a rectangular image frame composed of a pixel signal and a shape signal corresponding to the pixel signal as an input, and within a VOP (a region of a pixel signal having a shape by the shape signal). Is output as a bit stream, and V is used as a reference image for motion estimation processing.
The present invention relates to a moving picture coding method and apparatus including a reference padding process for substituting pixel values outside OP.

[0002]

2. Description of the Related Art In a moving image coding algorithm such as MPEG-4, a pixel value outside the VOP of a reference image is generated from a pixel value inside the VOP, so that a motion vector designating outside the VOP can also be generated. For this reason, reference padding for the reference image is required in the moving image encoding device and the decoding device.

In the reference padding process, the motion-compensated image is padded. Therefore, in the conventional moving image coding system, a macroblock having pixels in a VOP is coded and then locally decoded in the VOP. Generally, an image is generated, and then padding processing is performed on macroblocks outside the VOP and on the VOP boundary. Therefore, it is necessary to store the input image in the memory, then read the shape signal from the memory, determine the reference padding processing method, and then read the pixel signal from the memory to perform the reference padding processing.

The padding of the reference image is carried out in macroblock units, and one of the following three kinds of padding is carried out depending on the relationship with the VOP (reference "Kenichi Miki / ed." M.
All of PEG-4 can freely encode various video and audio ", Industrial Research Board, published in September 1998, ISBN:
4-7693-1167-2 ").

1) Macroblock horizontal and vertical padding containing VOP boundaries 2) VOPs adjacent to macroblocks containing VOP boundaries
Outside block expansion padding, 3) Macroblock constant value padding outside VOP other than the above.

The padding of the reference image is performed only within the bounding box of the reference image, and the macroblock position, which is a padding unit, is (motion compensation and D
This is a macroblock (which performs CT coding). When referring to pixel values outside the bounding box in motion detection / compensation, treat them in the same way as when referring to pixel values outside the frame in frame-based coding, and make sure that the pixel values at the boundary inside the pounding box are horizontal. -Continuous in the vertical direction (in the case where the boundary is a pixel outside the VOP, it is assumed that the values of the boundary padding / extended padding / constant value padding are continuous).

1) VOP boundary padding The VOP boundary padding is performed in two steps: horizontal padding and vertical padding.

A) Horizontal padding Horizontal padding is a method in which pixels inside a VOP are horizontally copied to obtain pixel values of pixels outside the VOP. V
The pixel value outside the OP is the average value of the pixel values within the VOP that are closest to the right position and the fixed position of the same horizontal line in the macroblock. If a pixel value within the VOP exists only at the right position or fixed position of the macroblock, the pixel value is set as a pixel value outside the VOP, and if there is no VOP internal image system at either the right position or the fixed position. , Vertical padding.

B) Vertical padding In vertical padding, macroblocks that have been padded in the horizontal direction are subjected to the same processing as horizontal padding in the vertical direction. That is, the pixel value outside the VOP is the closest V at the upper and lower positions of the same vertical column in the macroblock.
It is the average value of the pixel value in the OP or the horizontal padding value.
When the pixel value in the VOP or the horizontal padding value exists only at the upper position or the lower position of the macro block, the pixel value is set as the pixel value outside the VOP. The vertical padding is performed only on the non-VOP pixels for which horizontal padding has not been performed.

2) When a macro block outside the expanded padding VOP has pixels inside the VOP in any of four adjacent macro blocks in the upper, lower, left, and right directions, the pixels of the macro block including the pixels inside the VOP. Performing padding is called extended padding. VOP among adjacent macroblocks in 4 directions
When there are a plurality of macroblocks including internal pixels, one is selected according to the following rules a) to d) to be the reference macroblock (the reference macroblock is selected in the order of decoding by software). , Was determined considering that the memory can be efficiently accessed).

A) If the left macroblock includes pixels in the VOP, the left macroblock is set as the reference macroblock.

B) If the condition of a) is not satisfied but the upper macroblock includes pixels in the VOP, the upper macroblock is set as the reference macroblock.

C) If the conditions of a) and b) are not satisfied, but the right macroblock includes pixels in the VOP, the right macroblock is set as the reference macroblock.

D) When the conditions of a), b) and c) are not satisfied, the lower macroblock is set as the reference macroblock.

In the extended padding, the pixel value at the boundary between adjacent reference macroblocks is copied in the horizontal or vertical direction. Extended padding is realized by repeatedly copying the pixel value of the rightmost pixel of the macroblock to the right.
When the reference block is the right macroblock, it is repeatedly copied to the left side, and when the upper macroblock and the lower macroblock are reference macroblocks, it is repeatedly copied to the lower side or the upper side, respectively.

If the boundary pixel of the reference macroblock is outside the VOP, the result of the boundary padding in 1) is taken as the pixel value of that pixel.

3) Constant value padding In a macro block outside the VOP of the reference image, which is not subjected to the extended padding, a constant value 128 (corresponding to gray of the color signal) is set as the pixel value of the macro block. And

[0018]

However, in the above-described conventional moving picture coding system, the following problems occur in the processing of determining the processing method of the reference padding.

1. After storing the input image in memory, the shape signal needs to be read from memory to determine how to handle the reference padding. In the image frame that becomes the reference image, SDTV is about 10 Mbps, HDTV
In this case, a memory bandwidth of about 60 Mbps is required.

2. When the reference padding processing method is not determined at the time of input, it is necessary to perform the padding processing after the reference image is generated, and it is necessary to read or store the pixel signal from the memory for the padding processing. This processing depends on the shape of the VOP. As an example, 20% of the image frame is a target block for boundary padding, 20% is a target block for extended padding, and 10%.
Assuming that is the target block for constant padding,
An image frame serving as a reference image requires a memory bandwidth of about 23 Mbps for SDTV and about 140 Mbps for HDTV.

As described above, the conventional method has a problem that the memory bandwidth becomes large.

It is an object of the present invention to provide a moving picture coding method and apparatus in which a memory bandwidth is reduced in moving picture coding for performing reference padding processing.

[0023]

In order to achieve the above object, the present invention takes a rectangular image frame composed of a pixel signal and a shape signal corresponding to the pixel signal as an input,
The area of the pixel signal where the shape exists due to the shape signal (hereinafter,
In the moving picture coding method, a bit stream whose inside is called VOP) is output, and a reference padding process for substituting pixel values outside the VOP as a reference image for motion estimation processing is input. Sometimes, a first step of determining a reference image padding processing method from an input shape signal, and a second step of inputting a pixel signal, a shape signal, and a reference padding processing method and encoding the VOP. And a moving picture coding method characterized by having.

Alternatively, the first of the above moving picture coding methods
In the stage, for each block obtained by dividing the image frame into a plurality of blocks,-(state 1) a block in which pixels inside the VOP and pixels outside the VOP are mixed (hereinafter, referred to as VOP boundary), depending on the relationship with the VOP. -(State 2) A block outside the VOP in which any one of adjacent blocks in the four directions (up, down, left and right) has a pixel in the VOP,-(state 3)
The third step of classifying into blocks other than the VOP other than the above, and the processing method of reference padding according to the above states,
-(State 1), VOP boundary padding for performing horizontal and vertical padding,-(State 2), extended padding for padding with pixels of any of adjacent four blocks in up, down, left, and right directions. In the case of (state 3), there is provided a fourth step of determining a constant value padding, and a moving picture coding method characterized by the above.

Alternatively, in the third and fourth steps of the above moving picture coding method, one set corresponding to the number of blocks in the horizontal direction is set in the memory, and a plurality of data ((hereinafter referred to as reference padding processing (Referred to as a data area for determining the method) is prepared, and the fifth step in which the reference padding method is a constant value padding as an initial setting and the shape value of the number of pixels in the horizontal direction of one block are input, and − (State 1) All pixels are within VOP,-(State 2) All pixels are outside VOP,-(State 3) Pixels within VOP and pixels outside VOP are mixed, The sixth step of classifying into a line and the line state based on the state of the line one line before (corresponding to (state of past line) below) corresponding to the line being updated in the input block Taken in 5 steps From the state described above, the method of processing the reference padding stored in the memory is performed. If (state a) (state of past line) is constant value padding and (state 1), the next state is boundary padded,・ (State b) (state of past line) is boundary padding, and
If (state 1), the next state is boundary padding,
(State c) (state of past line) is extended padding,
And if (state 1), the next state in the VOP,
(State d) (state of past line) is within VOP, and
If it is (state 1), the next state is in the VOP, and (state e) (state of the past line) is a constant value padding, and if it is (state 2), the next state is a constant value padding, If (state f) (past line state) is boundary padding and (state 2), next state is boundary padding, (state g) (past line state) is extended padding, and , (State 2), the next state is extended padding, · (state h) (state of past line)
Is in the VOP and is (state 2), the next state is the boundary padding, ... (state i) (state of past line)
Is a constant value padding and (state 3), the next state is boundary padding, and (state j) (past line state) is boundary padding and (state 3) State is boundary padding, (state k) (state of past line) is extended padding, and if (state 3), next state is boundary padding, (state l)
(Past line status) is within VOP, and (Status 3)
If it is, the 7th step of updating the next state to the boundary padding, and the 6th and 7th steps are repeated for the number of blocks in the horizontal direction for the number of vertical lines of the block. When it is determined that the padding is the boundary padding when processing the line of, the state of the adjacent upper, lower, left, and right blocks is changed from the state of the past block. In this case, in the VOP, if- (past block state) is a constant value padding, extended padding that copies the next state from the pixel value of the lower adjacent block,-(past block state) is the boundary padding Boundary padding: -If the state of the block adjacent to the left is-(past block state) within the VOP, then-in the VOP,-(past block state) In the case of value padding or extended padding that copies from the pixel value of the lower adjacent block, extended padding that copies the next state from the pixel value of the adjacent block to the right, − (past block state) is the boundary padding In the case of, the boundary padding, the state of the block adjacent to the bottom- (state of the past block)
Is a constant value padding, extended padding that copies the next state from the pixel value of the block adjacent to the top, the state of the block adjacent to the right- (state of the past block)
Is a constant value padding or extended padding copying from the pixel value of the upper adjacent block, the next state is updated to the value of extended padding copying from the pixel value of the left adjacent block And a moving picture coding method characterized by having:

Further, according to the present invention, a rectangular image frame composed of a pixel signal and a shape signal corresponding to the pixel signal is input, a bit stream encoded in the VOP is output, and a motion estimation process is performed. VO as reference image
In a moving picture coding device having a reference padding process for substituting a pixel value outside P, when inputting an image frame, first means for determining a processing method of the reference image padding from an input shape signal, and a pixel And a second means for inputting a signal, a shape signal, and a reference padding processing method, and encoding the inside of the VOP.

Alternatively, the first of the above moving picture coding devices
For each block obtained by dividing the image frame into a plurality of blocks in the above means, depending on the relationship with VOP,-(state 1) a block in which pixels inside VOP and pixels outside VOP are mixed (VOP boundary),-(state 2) ) A block outside the VOP is a VOP that is adjacent to one of the four blocks in the up, down, left and right
Block with pixels inside,-(state 3) V other than the above
The third means for classifying the block outside the OP and the method of processing the reference padding depending on each of the states are as follows :-( state 1)
, Then V for horizontal and vertical padding
OP boundary padding,-(state 2), extended padding that performs padding on pixels in any of four adjacent blocks in the four directions of up, down, left, and right .- (state 3), fixed padding And means for a moving picture coding device.

Alternatively, in the third and fourth means of the moving picture coding apparatus, a plurality of blocks for determining the reference padding processing method are set on the memory by the number of blocks in the horizontal direction as one set. Fifth means for preparing the data area and making the reference padding method a constant value padding as an initial setting and the shape value of the number of pixels in the horizontal direction of one block are input, and- (state 1) all pixels A sixth means for classifying into a state within the VOP, a state (-state 2) in which all pixels are outside the VOP, and a state (-state 3) in which pixels inside the VOP and pixels outside the VOP are mixed. , The reference padding stored in the memory from the state acquired by the fifth means is repeatedly performed in the line direction based on the previous line state (state of the previous line) corresponding to the line being updated in the input block. The processing method is as follows: If (state a) (past line state) is constant value padding and (state 1), the next state is boundary padding, and (state b) (past line state) is Boundary padding, and
If (state 1), the next state is boundary padding,
(State c) (state of past line) is extended padding,
And if (state 1), the next state in the VOP,
(State d) (state of past line) is within VOP, and
If it is (state 1), the next state is in the VOP, and (state e) (state of the past line) is a constant value padding, and if it is (state 2), the next state is a constant value padding, If (state f) (past line state) is boundary padding and (state 2), next state is boundary padding, (state g) (past line state) is extended padding, and , (State 2), the next state is extended padding, · (state h) (state of past line)
Is in the VOP and is (state 2), the next state is the boundary padding, ... (state i) (state of past line)
Is a constant value padding and (state 3), the next state is boundary padding, and (state j) (past line state) is boundary padding and (state 3) State is boundary padding, (state k) (state of past line) is extended padding, and if (state 3), next state is boundary padding, (state l)
(Past line status) is within VOP, and (Status 3)
If it is, the seventh means for updating the next state to the boundary padding, the sixth means and the seventh means are repeated for the number of blocks in the horizontal direction for the number of vertical lines of the block, and the end of the block. When it is determined that the padding is the boundary padding when processing the line of, the state of the adjacent upper, lower, left, and right blocks is changed from the state of the past block. The state of the upper adjacent block is- (the state of the past block) is within the VOP. In this case, in the VOP, if- (past block state) is a constant value padding, extended padding that copies the next state from the pixel value of the lower adjacent block,-(past block state) is the boundary padding Boundary padding: -If the state of the block adjacent to the left is-(past block state) within the VOP, then-in the VOP,-(past block state) In the case of value padding or extended padding that copies from the pixel value of the lower adjacent block, extended padding that copies the next state from the pixel value of the adjacent block to the right, − (past block state) is the boundary padding In the case of, the boundary padding, the state of the block adjacent to the bottom- (state of the past block)
Is a constant value padding, extended padding that copies the next state from the pixel value of the block adjacent to the top, the state of the block adjacent to the right- (state of the past block)
Is a constant value padding or extended padding for copying from the pixel value of an upper adjacent block, an eighth means for updating the next state to a value of extended padding for copying from the pixel value of the left adjacent block And a moving picture coding apparatus characterized by comprising:

The present invention also provides a recording medium having a program for causing a computer to execute the above moving picture coding method or a program for causing a computer to execute the above moving picture coding method.

In the present invention, by determining the processing method of the reference padding from the shape signal when the input image is stored in the memory, it is not necessary to store the shape signal in the memory and then read it from the memory. Reduce O rate.

[0031]

BEST MODE FOR CARRYING OUT THE INVENTION Embodiments of the present invention will be described in detail below with reference to the drawings.

FIG. 1 shows the configuration and processing of a first embodiment of the present invention.

In the figure, reference numerals 1 and 2 represent input pixel signals and shape signals. In the figure, 1 indicates a pixel signal composed of a luminance signal and a color difference signal, and 2 in the figure indicates a shape signal corresponding to the pixel signal. Reference numeral 3 in the drawing denotes a moving picture coding system, and 4 in the drawing shows an input unit of the moving picture coding system.
The pixel signals 1 and 2 in the drawing and the shape signal are input to the input unit 4 in the drawing, and the processing method of the reference padding is determined from the input shape signal by the first means. (Luminance signal, color difference signal), the shape signal 6 in the figure, and the reference padding processing method 7 in the figure are output.

Reference numeral 8 in the drawing indicates a second means, in which the pixel signal 5 in the drawing, the shape signal 6 in the drawing, and the reference padding processing method 7 in the drawing are inputted, VOPs are encoded, and Output the middle 9 bitstream.

FIG. 2 shows a second embodiment of the present invention, showing the configuration and processing of the first means of the input section 4.

Reference numeral 21 in the drawing denotes an input shape signal. Reference numeral 22 in the figure indicates a third means, and from the input shape signal,
(State 1) Block at VOP boundary, (State 2) Block outside VOP in which adjacent blocks in any of four directions of up, down, left and right have pixels within VOP, (State 3) Block outside VOP other than the above ,,. 23 in the figure
Indicates the state determined by the third means, and is the fourth in 24 in the figure.
Is input to the means of.

In the fourth means 24 in the drawing, the reference image padding method is set according to the input state, and in the case of (state 1), VO for performing horizontal and vertical padding is used.
P boundary padding, in the case of (state 2), it is determined that it is extended padding that performs padding with pixels in any of four adjacent blocks in the four directions of up, down, left, and right, and in the case of (state 3), it is fixed value padding. To output.

FIG. 3 shows a third embodiment of the present invention, and shows the configuration and processing of the third and fourth means.

Reference numeral 301 in the drawing indicates a fifth means, which is 3 in the drawing.
Reference numeral 02 indicates a data area for three sets, where the number of blocks in the horizontal direction is one set. The fifth means sets the constant value padding as the reference padding method for all data areas as an initial setting.

Reference numeral 303 in the drawing indicates a shape value corresponding to an input pixel. The shape values are input in raster order for each line. Therefore, the state determination 305 is performed for each line, and the determination processing for each line is repeated for the number of lines in the block, whereby the states of all blocks included in one block row are determined. In the figure, as an example, the case where the number of pixels in the horizontal direction of the block is 8 pixels is shown. Reference numeral 304 in the figure indicates a sixth means, and 3 in the figure
The shape value of 03 is input to 305 in the drawing, and it is determined whether the shape value corresponds to the states indicated by 306, 307, and 308 in the drawing. In the figure, 306 indicates (state 1) all pixels are V
In the figure, 307 indicates that all the pixels (state 2) are outside the VOP, and 308 in the figure indicates
(State 3) Indicates that pixels inside the VOP and pixels outside the VOP are mixed.

Reference numeral 309 in the figure indicates a seventh means, which is 3 in the figure.
Reference numeral 10 indicates a memory. The seventh means is the state of the line one line before corresponding to the state determined by the sixth means and the line being updated in the update processing target block shown in FIGS. The following update processing is repeatedly performed in the line direction based on the state.). When (state a) (state of past line) is constant value padding and (state 1), the next state is boundary padding, (state b) (state of past line) is boundary padding, And (state 1), the next state is boundary padding, (state c) (past line state) is extended padding, and (state 1), the next state is VOP
Of the following: If (state d) (state of past line) is within VOP and is (state 1), the next state is within VOP, and (state e) (state of past line) is a constant value If the padding is (state 2), the next state is padded with a constant value. If (state f) (state of past line) is boundary padding and (state 2), the next state is set. Boundary padding :-( state g) (state of past line) is extended padding and if (state 2) is extended padding,-(state h) (state of past line) is VOP If the state is (state 2), the next state is boundary padded. If (state i) (state of past line) is constant padding and (state 3), the next state is Boundary padding, (state j) (state of past line) Is the boundary padding and (state 3), the next state is the boundary padding, and (state k) (state of the previous line) is extended padding and (state 3), the next state Boundary padding: If (state 1) (state of past line) is within VOP and (state 3), then padding next state.

The above processing is repeated for the number of lines in one block row to determine the reference image padding type of all blocks included in one block row. In FIG. 3, as an example, the data 311 in the figure is updated.

In the figure, reference numeral 312 denotes an eighth means, which repeats the sixth means and the seventh means and repeats for the number of blocks in the horizontal direction for the number of vertical lines of the block, and the last line of the block. When it is determined to be the boundary padding when processing, the state of adjacent upper, lower, left, and right blocks is changed from the state of the past block, and the state of the upper adjacent block is- (the state of the past block) is within the VOP, VOP
Among them, when- (past block state) is constant value padding, extended padding that copies the next state from the pixel value of the lower adjacent block, -when (past block state) is boundary padding,
Boundary padding, -If the state of the block adjacent to the left is- (past block state) is within the VOP, VOP
In the case of- (past block state) padding with a constant value or extended padding that copies from the pixel value of the lower adjacent block, extended padding that copies the next state from the pixel value of the right adjacent block ,-(State of past block) is boundary padding,
Boundary padding: -The state of the block adjacent to the bottom-If the state of the block in the past is a constant value padding, the extended padding that copies the next state from the pixel value of the block adjacent to the top-Adjacent to the right If the state of a block is padding with a constant value- (past block state), or extended padding that copies from the pixel value of an adjacent block above, the next state is copied from the pixel value of the block adjacent to the left. Update to the value of extended padding.

When the processing of the last block of one block sequence is completed, as shown by 313 in the figure, data indicating the type of reference image padding for the upper block sequence for which the process is completed is stored in the memory. , The next block is processed.

It should be noted that a part or all of the functions of the respective embodiments shown in FIGS. 1 to 3 are configured by a computer program, and the program is executed by the computer to realize the present invention. It is needless to say that the steps of each embodiment can be configured by a program of a computer and the program can be executed by the computer. A program for executing the processing steps, a computer-readable recording medium,
For example, FD (floppy disk (registered trademark)),
It is possible to record on MO, ROM, memory card, CD, DVD, removable disk, etc., and save or distribute. It is also possible to provide the above program through a network such as the Internet or electronic mail.

[0046]

As is apparent from the above description, according to the present invention, the processing method of reference padding can be determined at the time of inputting the shape signal. Therefore, it is necessary to store the shape signal in the memory and then read it from the memory. It is possible to reduce the I / O rate of the memory.

[Brief description of drawings]

FIG. 1 is a diagram showing a first embodiment example of the present invention.

FIG. 2 is a diagram showing a second embodiment example of the present invention.

FIG. 3 is a diagram showing a third exemplary embodiment of the present invention.

FIG. 4 is a diagram showing how area data is updated in the third embodiment.

[Explanation of symbols]

1 ... Pixel signal 2 ... Shape signal 3 ... Moving picture coding system 4 ... Input section 5 ... Pixel signal 6 ... Shape signal 7 ... Reference padding processing method 21 ... Shape signal 22 ... Third means 23 ... State 24 ... Fourth means 25 ... Reference padding processing method 301 ... Fifth means 302 ... Data area 303 ... Shape value 304 ... Sixth means 305 ... Judgment of state 306 ... (State 1) 307 ... (State 2) 308 ... (State 3) 309 ... Seventh means 310 ... memory 311 ... Block to be updated 312 ... Eighth means 313 ... Processing of the last block

   ─────────────────────────────────────────────────── ─── Continued front page    F-term (reference) 5C059 KK25 MA05 MB02 MB03 MB04                       MB16 MB27 NN03 NN28 PP04                       PP25 PP28 SS06 SS20 TA00                       TB07 TC34 TD13 UA02 UA38                       UA39

Claims (8)

[Claims] 1. A rectangular image frame composed of a pixel signal and a shape signal corresponding to the pixel signal is input,
The area of the pixel signal where the shape exists due to the shape signal (hereinafter,
In the moving picture coding method, the inside of which is referred to as VOP) is output as a bit stream, and a reference padding process for substituting pixel values outside the VOP as a reference image for the motion estimation process is input. Sometimes, a first step of determining a processing method of reference image padding from an input shape signal; a second step of inputting a pixel signal, a shape signal, and a processing method of reference padding, and encoding in a VOP; A moving picture coding method comprising: 2. In the first step, for each block obtained by dividing an image frame into a plurality of blocks, depending on the relationship with VOP,-(state 1) block in which pixels inside VOP and pixels outside VOP are mixed ( Hereinafter, referred to as a VOP boundary),-(state 2) a block outside the VOP, in which any one of adjacent blocks in four directions of up, down, left, and right has a pixel within the VOP,-(state 3) outside the VOP other than the above A third step of classifying into blocks, and a method of processing reference padding according to each of the states, in the case of-(state 1), VOP boundary padding for performing horizontal and vertical padding, and- (state 2). In the case of extended padding that performs padding with pixels in any of four adjacent blocks in the four directions of up, down, left, and right, and- (state 3), constant padding , The moving picture coding method according to claim 1, characterized in that it comprises a. 3. In the third and fourth steps, one set corresponding to the number of blocks in the horizontal direction is set on the memory,
A fifth step of preparing a plurality of relevant data (hereinafter referred to as a data area for determining a reference padding processing method) and making the reference padding method a constant value padding as an initial setting; The shape value of the number of pixels in the horizontal direction is input, and- (state 1) all pixels are in VOP,-(state 2) all pixels are out of VOP,-(state 3) in VOP The sixth stage of classifying into a state in which pixels and pixels outside the VOP are mixed, and the state of the line one line before corresponding to the line being updated in the input block (hereinafter, (state of past line )) Is repeated in the line direction, and from the state acquired in the fifth stage, the reference padding processing method stored in the memory is set as follows: (state a) (state of past line) is a constant value padding , , (State 1), the next state is boundary padding, (state b) (past line state) is boundary padding, and if (state 1), the next state is boundary padding, If (state c) (state of past line) is extended padding and (state 1), the next state is VOP
Of the following: If (state d) (state of past line) is within VOP and is (state 1), the next state is within VOP, and (state e) (state of past line) is a constant value If the padding is (state 2), the next state is padded with a constant value. If (state f) (state of past line) is boundary padding and (state 2), the next state is set. Boundary padding :-( state g) (state of past line) is extended padding and if (state 2) is extended padding,-(state h) (state of past line) is VOP If the state is (state 2), the next state is boundary padded. If (state i) (state of past line) is constant padding and (state 3), the next state is Boundary padding, (state j) (state of past line) Is the boundary padding and (state 3), the next state is the boundary padding, and (state k) (state of the previous line) is extended padding and (state 3), the next state If the (state 1) (state of the past line) is in the VOP and is (state 3), the next state is updated to the boundary padding, the seventh stage and the sixth stage When the last line of the block is processed and the boundary padding is determined, the states of adjacent upper, lower, left and right blocks are pasted by repeating the above and the seventh step for the number of blocks in the horizontal direction for the number of vertical lines of the block. From the state of the block of the above-If the state of the block adjacent to the above- (state of the past block) is within the VOP,
Among them, when- (past block state) is constant value padding, extended padding that copies the next state from the pixel value of the lower adjacent block, -when (past block state) is boundary padding,
Boundary padding, -If the state of the block adjacent to the left is- (past block state) is within the VOP, VOP
In the case of- (past block state) padding with a constant value or extended padding that copies from the pixel value of the lower adjacent block, extended padding that copies the next state from the pixel value of the right adjacent block ,-(State of past block) is boundary padding,
Boundary padding: -The state of the block adjacent to the bottom-If the state of the block in the past is a constant value padding, the extended padding that copies the next state from the pixel value of the block adjacent to the top-Adjacent to the right If the state of a block is padding with a constant value- (past block state), or extended padding that copies from the pixel value of an adjacent block above, the next state is copied from the pixel value of the block adjacent to the left. The eighth step of updating to the value of the extended padding, and the moving picture coding method according to claim 2. 4. A rectangular image frame composed of a pixel signal and a shape signal corresponding to the pixel signal is input,
In a moving picture coding apparatus provided with a reference padding processing for outputting a bitstream coded in a VOP as an output and substituting pixel values outside the VOP as a reference picture for motion estimation processing, when inputting an image frame, input And a second means for inputting the pixel signal, the shape signal, and the reference padding processing method, and encoding the inside of the VOP. A moving image encoding device characterized by the above. 5. In the first means, for each block obtained by dividing an image frame into a plurality of blocks, depending on the relationship with VOP,-(state 1) a block in which pixels inside VOP and pixels outside VOP are mixed ( VOP boundary),-(State 2) Block outside VOP, which is adjacent to any block in four directions of up, down, left, and right, and has a pixel within VOP,-(State 3) Block outside VOP other than the above According to the third means and each of the states, the reference padding processing method is :-( state 1), VOP boundary padding for performing horizontal and vertical padding ;-( state 2), adjacent Expansion padding for padding pixels in any of four blocks in the four directions of up, down, left, and right, and fourth means for determining to- (state 3) constant padding. The video coding device according to claim 4, characterized in that. 6. The third and fourth means, wherein one set corresponding to the number of horizontal blocks is set on the memory,
Fifth means for preparing a plurality of relevant data areas for determining the reference padding processing method and making the reference padding method a constant value padding as an initial setting; and a shape value of the number of pixels in one block in the horizontal direction , (State 1) all pixels are within the VOP,-(state 2) all pixels are outside the VOP,-(state 3) pixels inside the VOP and pixels outside the VOP are mixed. 6th means for classifying into the current state and the line direction based on the previous line state (the state of the past line) corresponding to the line being updated in the input block, and the 5th means. From the obtained state, refer to the processing method of reference padding stored in the memory. ・ If (state a) (state of past line) is constant value padding and (state 1), the next state is boundary padded. When (state b) (state of past line) is boundary padding and (state 1), the next state is boundary padding, (state c) (state of past line) is extended padding, and , (State 1), the next state is VOP
Of the following: If (state d) (state of past line) is within VOP and is (state 1), the next state is within VOP, and (state e) (state of past line) is a constant value If the padding is (state 2), the next state is padded with a constant value. If (state f) (state of past line) is boundary padding and (state 2), the next state is set. Boundary padding :-( state g) (state of past line) is extended padding and if (state 2) is extended padding,-(state h) (state of past line) is VOP If the state is (state 2), the next state is boundary padded. If (state i) (state of past line) is constant padding and (state 3), the next state is Boundary padding, (state j) (state of past line) Is the boundary padding and (state 3), the next state is the boundary padding, and (state k) (state of the previous line) is extended padding and (state 3), the next state Boundary padding: (state 1) (state of past line) is within VOP and is (state 3), seventh means for updating the next state to boundary padding, and sixth means And the seventh means are repeated for the number of blocks in the horizontal direction for the number of vertical lines of the block, and when the last line of the block is processed and the boundary padding is determined, the states of the adjacent upper, lower, left and right blocks are pasted. From the state of the block of the above-If the state of the block adjacent to the above- (state of the past block) is within the VOP,
Among them, when- (past block state) is constant value padding, extended padding that copies the next state from the pixel value of the lower adjacent block, -when (past block state) is boundary padding,
Boundary padding, -If the state of the block adjacent to the left is- (past block state) is within the VOP, VOP
In the case of- (past block state) padding with a constant value or extended padding that copies from the pixel value of the lower adjacent block, extended padding that copies the next state from the pixel value of the right adjacent block ,-(State of past block) is boundary padding,
Boundary padding: -The state of the block adjacent to the bottom-If the state of the block in the past is a constant value padding, the extended padding that copies the next state from the pixel value of the block adjacent to the top-Adjacent to the right If the state of a block is padding with a constant value- (past block state), or extended padding that copies from the pixel value of an adjacent block above, the next state is copied from the pixel value of the block adjacent to the left. Eighth means for updating to the value of the extended padding.
The moving picture coding device according to. 7. A program for causing a computer to execute the moving picture coding method according to claim 1. 8. A recording medium recording a program for causing a computer to execute the moving picture coding method according to claim 1.