JPH0795536A - Device and method for reversely reproducing moving image - Google Patents

Abstract

PURPOSE:To reversely reproduce moving images with the less capacity of a display memory by providing storage capacity just for pictures corresponding to the types of pictures, successively storing the pictures sent from an adding part, reading the pictures according to the order and sending them to a display device. CONSTITUTION:When the type of a picture sent to an adding part 167 is a P picture, for example, since it is the picture depending on an I picture or the P picture positioned before it, the adding part 167 conducts addition processing by using the I picture or P picture previously stored in a motion correction part 165 and after motion correction is performed, the picture is outputted to a display buffer 169 and the motion correction part 165. Thus, since the pictures sent to the display buffer 169 are sent to the display device in a latter step in a fixed order, they are displayed in the proper picture order as expressed in time reference information.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a moving image reverse reproducing apparatus and method for performing reverse reproduction using a digital moving image read from a recording medium.

[0002]

2. Description of the Related Art Recently, a plurality of types of digital data, for example, video data and audio data, are combined and recorded on a recording medium such as a CD-ROM, and these data are read and output as needed. Multimedia personal information equipment is planned and developed. This kind of equipment decodes the composite data read from the recording medium into data of each type, and performs a necessary decoding for each type of data.
G (Moving Picture coding Image Experts Group) has established various standards, which include compression / expansion methods for image signals and audio signals.

A basic decoding system based on this MPEG is shown in FIG. In the figure, the digital data recording medium 1
Is composed of, for example, a CD-ROM or the like, and records a plurality of types of digital data such as image data and audio data in a composite form. The composite data recorded on the digital data recording medium 1 is read by a pickup (not shown) and sequentially sent to the system decoder (host CPU) 3.

The system decoder 3 extracts the time information by decoding the composite data and outputs it to the clock controller 5, and also outputs the audio data to the audio decoder 7 and the image data to the image decoder 9, respectively. The clock controller 5 supplies a system clock to the audio decoder 7 and the image decoder 9 based on the time information from the system decoder 3.

The audio decoder 7 and the image decoder 9 transmit and receive time information to and from the clock controller 5 to perform decoding of audio data and image data in synchronization with each other, and obtain the obtained data respectively. It is sent to the output device 8 and the image output device 10.

In such a decoding system, a plurality of types of digital data, for example, audio data and image data are multiplexed based on time information from the data flow read from the digital data recording medium 1 by the system decoder 3. In order to serve as an input to the audio decoder 7 and the image decoder 9, the data streams of audio and video are distributed and decoded.

The syntax of a bit stream according to the MPEG standard is a sequence header and six layers (from top to bottom: sequence, GOP (Group).
ofPictures), pictures, slices, macroblocks (16 × 16 pixel blocks, which are units for calculating motion compensation vectors), and blocks (8 × 8)
It becomes the unit of calculation of DCT composed of elements)). A 32-bit byte-arranged start code is prepared. These bit patterns do not occur in the video bitstream except at the start. Start codes other than sequence error codes (which represent uncorrectable errors from storage media) are defined in the syntax. In front of the start code, an arbitrary number of 0s are added to keep the byte arrangement. The sequence header specifies the screen format and so on.
To enable playback from the middle, the sequence header can be added to the beginning of all GOPs.
The sequence header attached to OP can change only the quantization matrix. The contents after the sequence start code are the horizontal size (12 bits), vertical size (12 bits), pixel aspect ratio (4 bits), picture rate (4 bits), bit rate (18 bits), VBV (video) of the image. buffeting ve
riffer) buffer size (10 bits), limit parameter flag (1 bit), two quantization matrix flags (1 bit) and contents (64 × 8 bits).

GOPs are not completely independent. The first GOP in the sequence allows starting with an I-picture in original screen order, but at the beginning of a general GOP is M
-1 (M is a P-picture period, usually about 3 to 6) B-pictures. The some B-pictures can use predictions from the last I, P-picture of the previous GOP. For reproduction from the middle, in a GOP with a sequence header, it is necessary to discard the first M-1 B-pictures when reproducing from the middle. Closed G is included in the GOP layer.
OP flag and broken link (Broken l
ink) flag is provided. Closed GOP
The flag indicates that the GOP does not need the screen of the previous GOP. The first M-1 B-pictures of the GOP mean that only backward prediction is used. This is a flag set during encoding. Closed GOP
When a non-null sequence is edited in GOP units, a broken link flag for the GOP next to the joining portion must be set. The decoder must discard the first M-1 B-pictures of the GOP having the broken link flag, without displaying them. After the group start code, time code (25 bits), closed
GOP (1 bit), broken link flag (1 bit), etc. follow.

Pictures have four types that are functionally different. The D-picture is image data of only the DC component used for high speed feed and high speed reverse. The D-picture is a decoder essential requirement, but the D-picture is included in another sequence. The sequence is different from the normal image of I, P, B-pictures. After the picture start code, the temporal reference (10 bits), the picture code type of the picture type (3 bits), and the value of the input virtual buffer of the decoder are shown. After vbv (16 bits), the motion vector is an integer unit, and the required number of motion vector frame intervals is described for the P and B types. Other than the sequence and GOP header, extension and user data may follow.

A slice is a 16-pixel wide band of arbitrary length and cannot extend over a picture. The first and last macroblocks are non-skip (Nonskippe
d), and at least one macroblock is non-skip. After the slice start code including the vertical position of the slice, the quantization scale (5 bits) may be followed by the extra information.

Macroblocks do not have a start code. After MBA comes the macroblock type. Depending on the type, the quantization scale may follow. Also, the required number of motion vectors is attached. In Coded types other than intra, a macroblock pattern (CBP) is attached.

The block layer is a DCT (Discr).
ete Cosine Transform) coefficient is E
Continue to OB. The block dropped due to CBP does not have the block layer itself.

Generally, the sequence of image information is divided into several packets, and each packet is a GOP consisting of a plurality of pictures (one still screen) and lower layer data.
It is composed of a unit called "Group of Pictures". The sequence information is the top layer that represents the coding hierarchy, and includes the header and some GOs.
It consists of P. One GOP is a unit that can be randomly accessed as it is. This GOP is a minimum coding unit that can be independently decoded within the range of a sequence, and is composed of one header and a plurality of pictures. One picture corresponds to one frame of a moving image or movie. FIG. 10 shows G defined by the MPEG standard.
Three types of pictures that compose an OP, that is, I,
It shows the mutual relationship between P and B pictures, and each one picture corresponds to one frame of a moving image or movie. That is, I picture: This I picture is a random access picture that has the most data amount among the three types of pictures and is only subjected to basic data compression, and can be coded regardless of other pictures. Is. That is, it is an intra-frame (Intra frame) encoded screen, and all macroblocks are intra-frame encoded (Intra encoded). The purpose is to easily maintain the independence of GOPs.

P picture: This P picture is coded for motion compensation of the previous I or P picture. That is, it is an inter-frame (Predictive) predictive coding screen, and is CCITT H.264. Intra coding and interframe coding (In
(ter coding) is a screen type that can be selected.

B picture: This B picture is a picture with the highest data rate compression rate among the three types of pictures, and is coded for motion compensation of the preceding and following I or P pictures. That is, bidirectional
This is a screen type which is a predictive coding screen and is a screen type peculiar to MPEG, and not only the past I and P pictures can be used for prediction but also the future I and P pictures can be used for prediction.

As shown by the arrows in FIG. 10, the above-mentioned dependency relationship between I, P and B pictures is such that only I picture exists independently and P picture is located in front of I or P picture.
The B picture depends on the preceding and subsequent I or P pictures.

In some applications, there is a demand to reproduce the image signal in the reverse direction. However, since there are pictures generated depending on the previous or next pictures as described above, a memory having a capacity capable of storing all the pictures forming one GOP is used, and once the GOP is forwarded. If the decoder configuration is such that the pictures in the memory are reproduced and stored in the memory and then the pictures in the memory are read in the reverse direction, the above requirements cannot be realized. Therefore, a large amount of display memory is required to store the decoded picture as compared with the case of reproducing the image in the forward direction. As a result, it was practically difficult to realize the reverse reproduction as described above.

Further, if the reverse reproduction is performed by using only the I picture which is an independent picture which does not depend on other pictures in the GOP, the problem of the capacity of the display memory can be avoided, but on the other hand, Since the number of pictures actually used for display is greatly reduced, it is not practical because only a moving image with rough motion and no smoothness and low display quality can be obtained.

[0019]

As described above, when an image signal is reproduced in the reverse direction, there is a picture that is generated depending on the previous or next picture, and one G
A memory having a capacity capable of storing all the pictures forming the OP is used, the picture in the GOP is once reproduced in the forward direction and stored in the memory, and then the picture in the memory is read in the reverse direction. The demand as described above cannot be realized unless the decoder structure is set. Therefore, a large amount of display memory is required to store the decoded picture as compared with the case of reproducing the image in the forward direction. I can't afford
As a result, it was practically difficult to realize the reverse reproduction as described above.

Further, if the reverse reproduction is performed by using only the I picture which is an independent picture that does not depend on other pictures in the GOP, the problem of the capacity of the display memory can be avoided, but on the other hand, Since the number of pictures actually used for display is greatly reduced, it is not practical because only a moving image with rough motion and no smoothness and low display quality can be obtained. An object of the present invention is to provide a moving image reverse reproducing apparatus and method capable of realizing reverse reproduction of a moving image with a small display memory capacity.

[0021]

According to the present invention, a moving image in which image data decoding is performed in units of a group of compressed image data composed of a plurality of types of pictures having different degrees of dependence on other pictures in the group. In the image reverse reproducing apparatus, table creating means for creating a table showing forward display order information of each picture forming a group, picture type and byte position information within the group; and at the time of creating the table by the table creating means, Byte counting means for counting byte position information;
Table storing means for storing the table created by the table creating means; deciding means for deciding the decoding order of the pictures in the group during backward reproduction according to the contents of the table stored in the table storing means; A decoding means for decoding the pictures in the decoding order obtained by the deciding means; a data expanding means for expanding the picture decoded by the decoding means into a picture having a constant data amount corresponding to the picture type. And display storage means for sequentially storing the pictures obtained by the data expansion means.

Further, according to the present invention, a moving image reverse reproducing method for decoding image data in units of a group of compressed image data composed of a plurality of types of pictures having different degrees of dependence on other pictures in the group. In a), a table is created which represents the display order information in the forward direction of each picture forming the group, the picture type, and the byte position information in the group, and b) the byte position information is counted when the table is reproduced, and c ) Store the table created above, d) determine the decoding order of the pictures in the GOP during backward playback according to the contents of the stored table, and e) execute the decoding of the pictures in the determined decoding order. F) Decoding the decoded picture with a fixed amount of data corresponding to the picture type. It extends catcher, g) executes display the extended picture sequentially stored to.

[0023]

With the above configuration, it is possible to greatly reduce the storage area required to store the type of picture that has the largest number in the GOP and has a high degree of dependence on other types of pictures. Therefore, reverse reproduction of a moving image can be realized with a small display memory capacity.

[0024]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the moving image reverse reproducing apparatus and method of the present invention will be described below. The terms used in the following description and their definitions are shown below. Definition: Inverse Motion Vector: A motion vector that is later used to motion compensate a reference picture in display order.

Bidirectional predictive coded code; B-picture:
An image (picture) encoded with motion-compensated prediction from past and / or reference pictures. Coded audio bitstream: ISO / IEC
A series of coded images defined in 11172.

DCT coefficient: Magnitude of specific cosine basic function Decoding: A process defined by ISO / IEC11172, which reads an input coded bitstream and outputs a decoded picture or audio sample.

Dequantization: The process of rescaling the quantized coefficients before the quantized bitstream is decoded and provided to the inverse DCT section. DCT (Discrete Cosine Trans)
form): Effective orthogonal transform used for image coding.

High-speed forward reproduction: A sequence of images or a part of a sequence is displayed in display order faster than real time. Forward motion vector: A motion vector used for motion compensation from a reference picture earlier in the display order.

Frame: A portion of the audio signal corresponding to audio PCM samples from the audio access unit. GOP (group of pictures): A series of coded pictures to assist random access. GOP is one of the coding syntax layers defined in ISO / IEC11172.

Intra-coding: Coding of a macroblock or picture using only information from the macroblock or picture. Macroblock: 4 blocks of luminance signal (Y) and 2 blocks of color difference signal (Cb, Cr). Each block is 8x
It consists of 8 pixels. 4 blocks of luminance and 1 block of color difference have the same size on the screen. Macroblocks are pixel data, codes that represent pixel values, and ISO / IEC111
It may also be used to call other data elements defined in the macroblock layer of the syntax defined in 72.

Motion compensation: Improving pixel value prediction efficiency using motion vectors. Prediction uses motion vectors to offset past and / or future reference pictures that include previously decoded pixel values used to form the prediction error signal.

Motion vector: A two-dimensional vector used for motion compensation that gives an offset between the coordinate value of the current picture and the coordinate value of the reference picture. Non-Intra Coding: Coding of a macroblock or picture with information from both itself and the macroblock and picture that occur at different times.

Predictive coded picture; P-pictur
e: Picture coded using motion compensated prediction from past reference pictures. Prediction error: The difference between the actual value of a pixel and the predicted value.

Quantization matrix: 64 8-bit values used by inverse quantization. Quantized DCT coefficient: DCT coefficient before dequantization. Variable length coded quantized DCT coefficients are stored as part of the compressed video bitstream.

Reference picture: In display order,
The I or P picture that is most adjacent to the current picture. Variable length coding; VLC: A reversible coding that assigns shorter codewords to frequent events and longer codewords to infrequent events.

FIG. 1 is a detailed block diagram of the video decoder of the present invention. The video decoder 151 shown in FIG.
Input buffer 153, variable length code (VLC) decoder 155, dequantizing unit (dequantizing se)
action) 157, inverse DCT (Inverse Di)
screen Cosine Transform) section 159, decoding scan table storage section 161, byte counter 163, motion correction section 165, addition section 16
7, display buffer 169, and internal controller 17
It consists of 1. The internal controller 171 actually has a CPU and a control program, and integrally controls the other circuits described above.

The VLC decoder 155 outputs D in scan order.
The CT coefficients are arranged, and VLC is assigned to a set of leading zeros (runs) and quantized coefficient values (levels).

The inverse DCT unit 159 has an N × N two-dimensional DCT (DCT (Discrete) expressed by the equation (1).
(Cosine Transform) is an effective orthogonal transform used for image coding, and is composed of well-known dedicated hardware), and the amount of information reduced in the spatial direction of the image is calculated using Equation (2). To restore the original amount of information.
However, xi, j represents a real space value, and Xk, l represents a two-dimensional DCT coefficient (integer).

[0039]

[Equation 1]

[0040]

[Equation 2]

The bit stream of the distributed video data sent from the host CPU is accumulated in the input buffer 153 every 1 GOP. For this input buffer 153, the VLC decoder 155 decodes only the header part of each picture in the GOP and extracts each temporal reference information and each picture type.

A byte counter 163 is additionally provided to the VLC decoder 155. This byte counter 16
3 corresponds to a so-called pointer and is for knowing how many bytes from the head position in the GOP there is a necessary picture, and its count value is "0" each time a bit stream of each new GOP is reached. Is reset to, and continues to count up sequentially until the next reset.

The VLC decoder 155 forwards the header of each picture in the forward direction (forward direction).
n), the decoding scan table is created and stored in the decoding scan table storage unit 161 during reproduction in the forward direction, for example. The VLC decoder 155 creates this decoding scan table while performing the decoding of each picture. The picture to be decoded is transferred to the VLC decoder 155 according to the decoding scan table stored in the decoding scan table storage unit 161, and is decoded according to the type of picture in a new order.

Since a B picture depends on the next I or P picture (in display order), the I or P picture must be transmitted and decoded before its dependent B picture. A picture consists of a header and one or more slices. The picture header contains the time, picture type,
And coding information is included. Slices are provided as a measure when data is corrupted. If the bitstream of a picture becomes unreadable,
The decoder can recover without discarding the entire picture by waiting for the next slice.

The input bitstream is stored in the input buffer until needed. The variable length decoder 155 decodes the header of the first picture, picture 0, and
Judge as a picture. The variable length decoder 155 generates a quantized coefficient corresponding to the quantized DCT coefficient. These quantized coefficients are converted to 8 in the image by the inverse zigzag scan.
Each x8 pixel block is assembled. (Prediction error is divided into blocks and DCT-transformed to obtain 8x8 DCT
Change the coefficients in a one-dimensional order using a zigzag scan). The inverse quantization circuit 157 uses the quantization step size to generate the actual DCT coefficient. The DCT coefficient is then converted into a pixel value by the inverse DCT conversion circuit 159. Pictures can be displayed at any suitable time.

The picture 3 which is the next picture of the VLC decoder 155 is decoded to determine that it is a P picture. For each block, the VLC decoder 155 decodes the motion vector that replaces the previous picture and the quantized coefficient corresponding to the quantized DCT coefficient of the difference block. These quantized coefficients are dequantized to produce the actual DCT coefficients. The DCT coefficients are then converted to pixel difference values and added to the predictive block obtained by applying the motion vector to the previous picture. This picture cannot be displayed until B-pictures 1 and 2 have been received, decoded and displayed.

The VLC decoder 155 decodes the header of the next picture, picture 1, and determines that it is a B picture. For each block, the VLC decoder 155 decodes a motion vector that replaces a past picture, a future picture, or both, and also decodes a quantized coefficient corresponding to a quantized DCT coefficient of a difference block. These quantized coefficients are dequantized to produce the actual DCT coefficients. The DCT coefficients are then inverse transformed into differential pixel values and added to the prediction block obtained by applying the motion vector to the stored picture. The block obtained as a result is the display buffer 169.
Stored in. This block can be displayed at any suitable time.

The VLC decoder 155 decodes the header of picture 2, which is the next picture, and determines that it is a B picture. This decoding is done in the same way as picture 1. After the picture 2 is decoded, the picture 0 stored in the display buffer 169 is unnecessary and therefore discarded.

Then, the VLC decoder 155 decodes each picture and sends it to the inverse quantizer 157. The picture inversely quantized by the inverse quantization unit 157 is further deDCTed by the inverse DCT unit 159 to be expanded to the original data amount, and then sent to the addition unit 167.

The motion compensation unit 165 divides the input image into blocks of a certain size (16 × 16), prepares a motion vector for each block, and shifts only the motion vector from the previous reproduced image (predicted image) for prediction. The effect of reducing prediction error power is generally greater than simple interframe prediction. In this case, the motion vector information is sent as additional information of the block. The motion vector is (m in Equation (3) below.
vx, mvy), which is added to the coordinates of the block of the predicted image. The coordinates are positive in the right direction and positive in the downward direction. Since it is shifted from the predicted image, the coded block is not shifted to form a gap.

The code amount of the motion vector, which is the additional information, is MC (motion compensation).
It does not usually cause a problem in view of the reduction of the code amount of the contents.
Color difference images are generally 1/2 the size of luminance images.
Therefore, the motion vector is divided by two.

Here, i and j are subscripts of pixels, k is a subscript of the screen, differencei, j, k are images to be encoded, inputi, j, k are input images, and reconstrru
cti, j, k-1 represents a previously reproduced image.

[0053]

[Equation 3]

The motion vector is basically expressed in half pel (half pixel) units. (By specifying the picture layer, integer representation is also possible.) Half-pel motion compensation takes a rounded average of 2 pixel values if the position of the predicted pixel is between 2 pixels, and a rounded average of 4 pixel values if it is between 4 pixels. . Therefore, half-pel motion compensation is not only the implication of improving the prediction accuracy,
Acts as a spatial low-pass filter.

In the B-picture, bidirectional half-pel motion compensation is performed. In the case of an interpolating macroblock that uses both predictions before and after, the rounded average of the two motion compensation predictions before and after is used as the prediction value. In motion compensation, the motion vector (M
VLC encoding the difference of V). The difference between the motion vector of this macroblock (MB) and the motion vector of the immediately preceding macroblock is encoded. However, if the first macroblock of the slice and the previous one are not motion compensated, the difference is not used and the value is used as it is.

In the B-picture, there can be two motion vectors, and the difference is reset only at the slice head and intra. That is, the values are made continuous even if the prediction in the previous direction is not used.

When the value represented by VLC is other than 0, the number of remaining bits (FLC) determined by the frame interval shown in the picture layer is added after this.
(1 bit when the frame interval is 2, 2 bits when the frame interval is 4, 3 bits when the frame interval is 8) 3-bit forward_f_code (or backwa of the picture layer)
Depending on the rd_f_code) value from 1 to 7, the frame interval becomes 1, 2, 4, 8, 16, 32, 64,
Among them, 1, 2, 4, and 8 are within the limiting parameter.

The frame interval may be different from the actual frame interval. For example, in a P-picture with M = 3, the actual frame interval is 3, but since 3 is not prepared, 4
Can be used.

In the picture layer, full_p
el_forward_vector (B-pictu
In re, it and full_pel_backward-
vector) is prepared and when this is 1,
The motion vector is an integer unit.

The decoding of one component of the motion vector can be written, for example, by the following C language program.
(Get_flc (n) is from bitstream to n
A function that takes only bits, get_mv_vlc () is V
Decode_mvc (mvc, f_code) {int v, frame_dist, sgn, max, min, r; v = get_mv_vlc (); / * code other than motion_vector_forward_bit * code with bit * code_code * / If (== 0) return mvc; / * If the value is 0, it ends with the previous value * .sgn = get_flc (1); / * When the value is other than 0, the sign bit is taken. * / Frame_dist = 1 <<(f_code-1); r = 0; if (frame_dist> 1) r = get_flc (f_code-1); / * When the frame interval is 2 or more, the remaining FLC motion_horizonal_forward-r is taken. . * / Max = 16 * frame_dist-1; / * maximum value * / min = -16 * frame_dist; / * minimum value * / mvc + = sgn * ((v-1) * frame_dist + (r + 1); / * / * If (mvc <min) mvc + = 32 * frame_dist; / * smaller than the minimum value * / else if (max <mvc) mvc- = 32 * frame_dist; / * larger than the maximum value * / return mvc; * Returns the component of the motion vector * / The difference value is reproduced as described above with the value of VLC and the value represented by the following bit. This is because only 0 has no remaining bit.

The method of reproducing the motion vector component allows two interpretations for differential encoding, and after addition of the difference, the vector component value is −16 × frame interval (minimum value).
If it does not come within the 16x frame interval -1 (maximum value), the other side interprets and the 32x frame interval is added or subtracted. Therefore, the description range of the motion vector is in units of half pixels (or in units of integer pixels), when the frame interval is 1, +15 to -16, when +2, +31 to -32,
+4 to +64 for 4 and -127 for 8
To -128.

Quantization expresses the entire DCT coefficient by a numerical value with a small value, and by doing so, the code amount can be reduced. Therefore, the reproduction of the DCT coefficient is an approximate value. Quantization is performed with a stepsize that is twice as large as a quantum described in a slice or a macroblock, and Q described below.
Divide the coefficient 16 by the product of M.

Inverse quantization is performed by the following equations for both luminance and color difference, except for the intra DC coefficient. i = scan [m] [n]; / * zigzag scan * / In Intra, rec [m] [n] = 2 * level [i] * quant * in tra_quant [m] [n] / 16; In Non-intra rec [m] [n] = (2 * level [i] + sign (level [i]) * quant * non_intra_quant [m] [n] / 16; if ((rec [m] [n] & 1) == 0 ) Rec [m] [n] = rec [m] [n] -sign (rec [m] [n]); if (rec [m] [n]> 2047) rec [m] [n] = 2047; if (rec [m] [n] <− 2048) rec [m] [n] = − 2048; In Non-Intra, 2 * level is set to sign (level) to reproduce the image with the vicinity of 0 separated. ) Are added to. That way the inverse quantization means that use a simple division by quantization and releasing the vicinity of 0 may serve to suppress noisy changes in Non-intra.

Also, one or two pictures for motion correction are input to the adder 167 from the motion corrector 165 which holds the output of the adder 167 according to the type of picture. When a picture is input from the motion compensation unit 165, the addition unit 167 performs addition processing on a total of two to three pictures to obtain a new picture, and when no picture is input from the motion compensation unit 165, the inverse DCT unit 159 The picture is output to the display buffer 169 as it is. The display buffer 169 has a storage capacity of (total number of I and P pictures in 1 GOP + 2) pictures, sequentially stores the pictures sent from the adder 167, reads the pictures in the order described later, and reads the pictures. Send to display device.

Next, the operation of the above embodiment will be described separately for the case of performing forward reproduction and the case of performing reverse reproduction. When performing reproduction in the forward direction, it is assumed that the GOPs of the image data read from the host CPU to the image decoder 151 are in the order shown in FIG. This order is a decoding order during reproduction in the forward direction, and the temporal reference information in the figure is information indicating the order of pictures in the GOP when actually displayed, and is set in the header portion of the picture. include.

The VLC decoder 155 outputs GO to the image data sent via the input buffer 153.
The byte counter 163 is reset to the start position of P to start counting up the number of bytes from "0", and only the header part of each picture sent is decoded to obtain each time reference information and each picture type. And are retrieved via the appropriate header parameters.

The VLC decoder 155 receives the thus obtained temporal reference information, picture type and byte counter 163.
A decoding scan table as shown in FIG. 3 is created based on the count value of 1 and stored in the decoding scan table storage unit 161. This decoding scan table is used during reverse reproduction as described later. Note that the byte length of each picture attached to the right end in the figure is shown for reference and is not actually stored in the table.

FIG. 4 is a flow chart showing the procedure for creating the decoding scan table. The program shown by this flowchart is stored in the memory (ROM) in the internal controller 171 shown in FIG.

In FIG. 4, the internal controller 171
Receives the bitstream supplied from the system decoder (eg, host CPU) in step 173, and GOP_start_in step 175.
Judge whether it is a code or not. GOP_start_co
If it is determined to be de, the controller 171 resets the byte counter 163 in step 177, and counts the same counter 163 in step 179. Next, the internal controller 171 performs step 1
81, the byte data of the bit stream is Pi
It is determined whether or not it is the picture_Start_Code. If it is not Picture_Start_Code, the processing returns to step 179, and Picture_Start is started.
Steps 179 and 181 are repeated until _Code appears.

In step 181, the Picture
Upon detecting _Start_Code, the internal controller 171 determines in step 183 that the temporary
A table is created based on the reference coding type and the count value of the byte counter 163, the process returns to step 175, and steps 175 to 183 are repeatedly executed until the last byte of the bit stream is processed.

Further, the VLC decoder 155 creates the decoding scan table as described above during this forward operation, and at the same time, for each picture sent from the host CPU in the order shown in FIG. Decoding according to the length is performed and the result is sent to the inverse quantization unit 157.

The picture, which has been dequantized by the dequantization unit 157, decompressed by the inverse DCT unit 157 and expanded to the original data amount before compression, is sent to the addition unit 167.

Here, if the type of the picture sent to the addition section 167 is an I picture, it is a picture that does not depend on other pictures, so the addition section 167 does not perform any addition processing and inputs it. The I picture is output as it is to the display buffer 169 and the motion correction unit 165.

When the type of the picture sent to the adder 167 is the P picture, the I located before it is
Since the picture depends on the picture or the P picture, the addition section 167 executes the addition processing using the I picture or the P picture stored in advance in the motion correction section 165, and after the motion correction is performed, the display buffer 1
69 and the motion correction unit 165.

Furthermore, when the type of the picture sent to the addition section 167 is a B picture, it is a picture that depends on an I picture or a P picture located before and after it, so the addition section 167 has a motion compensation section 165 in advance. The addition process is executed by using the I picture or P picture stored in the above, and after the motion correction is performed, it is output only to the display buffer 169.

The pictures sent to the display buffer 169 in this way are sent to the display device in the subsequent stage in the order shown in FIG. 5, and as shown in the time reference information, they are in the correct picture order. It is displayed (in the forward direction).

Next, the operation during reverse reproduction will be described. At the start of the reverse reproduction operation, the internal controller 171 searches for an I picture or a P picture in the decoding scan table (FIG. 3) stored in the decoding scan table storage unit 161 from its head position. Therefore, first, when the time reference number "0" of the I picture located at the beginning is obtained, the corresponding byte count value "0" is read out and transferred to the host CPU side via the input buffer 21, and the GOP within the GOP concerned is read. It is requested to input the I picture of the time reference information "0" at the "0" byte.

In response to this request, when the system decoder 12 sends the I picture of the time reference information "0" at the "0" byte in the GOP, the VLC decoder 155 determines the byte length of the I picture. According to the above, and outputs to the inverse quantization unit 157.

Thus, the I picture of the time reference information "0" dequantized by the dequantization unit 157, decompressed by the inverse DCT unit 159 and expanded to the original data amount before compression is added. Sent to section 167.

The addition section 167 does not perform any addition processing on this I picture, and inputs the time reference information "0".
The I picture of is output to the display buffer 169 and the motion correction unit 165 as it is.

After that, the internal controller 171 determines the next I picture or P in the decoding scan table.
Search for a picture. Here, when the temporal reference number “3” of the P picture is obtained, the corresponding byte count value “600” is obtained.
00 "and reads the data from the host C via the input buffer 21.
Transfer to the PU side and request to input the P picture of the time reference information "3" at the "60000" th byte in the GOP.

When the P picture of the time reference information "3" at the "60000" th byte in the GOP is sent from the host CPU in response to this request, the variable length code decoder 22 makes the byte of the P picture concerned. Decoding according to the length is performed and the result is sent to the inverse quantization unit 157.

Thus, the P picture of the time reference information "3" dequantized by the dequantization unit 157, decompressed by the inverse DCT unit 159 and expanded to the original data amount before compression is added. Sent to section 167.

The adding section 167 has a time reference "0" stored in advance in the motion correcting section 165 for this P picture.
The addition processing is performed using the I picture of the above, and after the motion correction is performed, it is output to the display buffer 169 and the motion correction unit 165.

In this way, I pictures or P pictures are sequentially searched from the start position of the decoding scan table,
The corresponding byte count value is read out, transferred to the host CPU via the input buffer 153, and requested to input the I picture or P picture. And
Each time an I picture or a P picture is sent from the host CPU in response to these requests, the VLC decoder 155 performs decoding according to the byte length of the picture, and the inverse quantizing unit 157 and the inverse DCT unit 159 operate. Through the addition unit 167, addition processing is executed as necessary to obtain a displayable picture, and the display buffer 169 and the motion correction unit 1 are obtained.
To 65.

After that, when the decoding of the I picture or P picture is completed, the internal controller 17
1 then searches for a B picture from the end position of the decoding scan table. Therefore, first, when the time reference number "14" of the B picture located at the end is obtained, the corresponding byte count value "146470" is read and transferred to the host CPU via the input buffer 153, and the GOP
Time reference information “1” in the “146470” byte of
4 ”B picture is requested to be input.

In response to this request, when the B picture of the time reference information "14" at the "146470" th byte in the GOP is sent from the host CPU, the VLC decoder 155 sets the byte length of the B picture. Decoding is performed accordingly and the result is sent to the inverse quantization unit 157.

Thus, the B picture of the time reference information "14" which is dequantized by the dequantization unit 157, decompressed by the inverse DCT unit 159 and expanded to the original data amount before compression is added. Sent to section 167.

The addition unit 167 performs addition processing on this B picture using the I picture of the temporal reference information “12” and the P picture of the temporal reference information “15” stored in advance in the motion correction unit 165. It is executed and subjected to motion compensation, and then output only to the display buffer 169.

After that, the B picture is sequentially searched from the end position of the decoding scan table, the corresponding byte count value is read out, and the host CP is read via the input buffer 21.
Transfer to U and request to input the B picture. Then, each time a B picture is sent from the host CPU in response to these requests, the VLC decoder 155
Decoding according to the byte length of the picture with,
The addition unit 1 via the inverse quantization unit 157 and the inverse DCT unit 159.
At 67, an addition process is executed using the I picture or P picture stored in the motion correction unit 165 to obtain a displayable picture and send it to the display buffer 169.

Thus, when the decoding of the B picture is completed, the VLC decoder 155 completes the decoding for one GOP. Figure 6 is thus VL
The order of pictures decoded by the C decoder 155 is shown in a form corresponding to the decoding scan table. (It is not stored in the decoding scan table storage unit 161 in the state of FIG. 6.)
However, in the display buffer 169, a plurality of pictures to be stored are sent to the display device in the subsequent stage in the order as shown in FIG. 7, and as indicated by the time reference information, it is completely opposite to that in the forward reproduction. Are displayed in the order of.

FIG. 8 is a flow chart briefly showing the above-mentioned reverse reproduction operation. The internal controller 171 inputs the bitstream to the input buffer 15 in step 185.
Receive from 3. In step 189, the VLC decoder 155 decodes the bitstream for each picture according to the decoding order stored in the table.
Furthermore, the above-mentioned inverse quantization, inverse DCT processing, addition processing,
Performs motion correction processing. In step 191, the decoded and processed bit stream is stored in the display buffer 169. Internal controller 1
In step 193, the 71 displays the bitstream on the display device at the display timing. If not, return to step 189, step 18
9 to 193 are repeatedly executed.

Here, the storage capacity of the display buffer 169 will be described. In the display buffer 169, first, the adding unit 1
All I-pictures and P-pictures sent from 67 are stored. After that, when B pictures are continuously sent, the display buffer 169 uses the storage area for two pictures and while reading one picture, reads the other picture that has already been stored at the same time. The operation of sending to the display device is alternately repeated.

This will be described in the order of decoding shown in FIG. 6, for example. In the display buffer 169, after the P picture of the temporal reference information “15” is stored, when the B picture of the temporal reference information “14” is written in the first area for the B picture, the temporal reference information “ The P picture of "15" is read and sent to the display device.

After that, when the B picture of the time reference information "13" is sent, it is written in the second area for the B picture, and at the same time, the time reference information "1" of the first area which has already been written. The B picture of "14" is read and sent to the display device.

Next, when the B picture of the time reference information "11" is sent, the B picture is overwritten in the first area for the B picture, and at the same time, the time reference information "2" of the second area which has already been written. The B picture of "13" is read and sent to the display device.

Next, when the B picture of the time reference information "10" is sent, the B picture is overwritten in the second area for the B picture, and at the same time, the time reference information "12" of which the writing has already been completed. The I picture is read and sent to the display device.

After that, when the B picture of the time reference information "11" in the first area which has already been written is read and sent to the display device, the B picture of the time reference information "8" is sent. , The B picture of the time reference information "10" of the second area which has already been written is read out and sent to the display device.

As described above, while the B picture is written in one area, the operation of reading and displaying the other B picture already written in the other area is alternately repeated. The storage capacity required for the display buffer 169 is (total number of I and P pictures in 1 GOP + 2) pictures, and the required storage capacity can be significantly reduced.

Furthermore, P of the above-mentioned time reference information "15"
If the last picture in the GOP is an I picture or a P picture, like the picture, the display buffer 16
It is also possible to start reading and displaying of the I picture or P picture at the time when writing of B picture to 9 is started, so that the storage capacity of the display buffer 169 can be further reduced accordingly.

[0101]

As described above, according to the present invention, the GOP
It is possible to significantly reduce the storage area required to store the type of picture that has the largest number of pictures and has a high degree of dependence on other types of pictures. It is possible to realize the reverse reproduction of.

[Brief description of drawings]

FIG. 1 is a block diagram showing an embodiment of a moving image reverse reproducing apparatus of the present invention.

2 is an explanatory diagram showing a decoding order of pictures when reproducing video data in the forward direction in the embodiment shown in FIG. 1. FIG.

3 is a diagram showing an example of a scan table stored in a decoding scan table storage section shown in FIG.

FIG. 4 is a flowchart showing a procedure for creating a decoding scan table.

FIG. 5 is an explanatory diagram showing a display order of pictures when reproducing video data in the forward direction.

FIG. 6 is an explanatory diagram showing a decoding order of pictures when reproducing video data in the reverse direction.

FIG. 7 is a diagram showing a display order of pictures when video data is reproduced in the reverse direction.

FIG. 8 is an explanatory diagram showing a procedure of reproducing video data in the reverse direction using a scan table.

FIG. 9 is a block diagram showing the configuration of a general decoding system based on the MPEG standard.

FIG. 10 is a diagram showing a dependency relationship of pictures in a GOP based on the MPEG standard.

[Explanation of symbols]

151 ... Video decoder, 153 ... Input buffer, 155 ... Variable length code decoder, 157 ...
Inverse quantization unit, 159 ... Inverse DCT unit, 161, ... Decoding scanning table storage unit, 163 ... Byte counter, 165 ... Motion correction unit, 167 ... Addition unit, 169. ． ． Display buffer, 171 ... Internal controller

─────────────────────────────────────────────────── ─── Continuation of the front page (51) Int.Cl. ⁶ Identification code Internal reference number FI Technical indication H04N 7/13 Z

Claims (3)

[Claims] 1. A moving picture reverse reproducing apparatus for decoding image data in units of a group of compressed image data composed of a plurality of types of pictures having different degrees of dependence on other pictures in the group, thereby forming groups. Table creating means for creating a table showing forward display order information of each picture, picture type and byte position information within a group; and byte counting means for counting the byte position information when creating the table by the table creating means. Table storing means for storing the table created by the table creating means; deciding means for deciding the decoding order of the pictures in the group during backward reproduction according to the contents of the table stored in the table storing means; The decoding order obtained by this decision means And a data decompression unit for decompressing the picture decoded by the decoding unit into a picture having a constant data amount corresponding to the picture type; and the data decompression unit. A moving image reverse reproducing apparatus, comprising: a display storage unit for sequentially storing pictures. 2. The table includes temporal reference information indicating an order of pictures in a group when actually displayed.
2. The moving picture reverse reproducing apparatus according to claim 1, wherein the type of each picture and the count value of the byte counting means are stored in the decoding order of each picture. 3. A moving image reverse reproducing method for decoding image data in units of a group of compressed image data consisting of a plurality of types of pictures having different degrees of dependence on other pictures in the group, wherein A table representing the display order information in the forward direction of each constituent picture, the picture type, and the byte position information within the group is created, b) the byte position information is counted when the table is reproduced, and c) the created table. And d) determining the decoding order of the pictures in the group during backward playback according to the stored table contents, e) performing the decoding of the pictures in the determined decoding order, and f) the decoding. The expanded picture is expanded into a picture of a fixed amount of data corresponding to the picture type, g Moving picture reverse playback method characterized by executing the display the extended picture sequentially stored to.