JPH03125570A - Coding method for picture data - Google Patents

Abstract

PURPOSE:To attain efficient coding by coding a difference between a picture element data of a frame coded already and a frame succeeding to a desired frame. CONSTITUTION:Each signal element of frames such as FR10, FR90... belonging to the frame is divided into plural macro blocks and signal elements are averaged for each macro block to obtain a mean value of signals in the block and the in-block signal mean value of each macro block is coded to generate a reference data FUN. Moreover, a difference between each signal element belonging to each frame and the in-block signal mean value belonging to each macro block is obtained and the difference is coded to generate an in-frame difference data IDEL. Then as to frames succeeding to a desired frame, the difference from the picture element data of the frame coded already is coded. Thus, the coding is implemented efficiently.

Description

[Detailed description of the invention]

"Industrial Application Field" The present invention relates to an image data encoding method suitable for use when recording image data on a storage medium such as a CD-ROM. ``Prior Art'' In order to record image data on storage media with limited storage capacity, it is necessary to perform encoding efficiently to keep the number of bits of the code string to be recorded small. The interframe coding method utilizes the correlation between consecutive moving images on the time axis. When encoding a frame's worth of image data, the difference between the image data of the immediately preceding encoded frame is encoded. According to this interframe encoding method, encoding can be performed using a much smaller number of bits than when encoding image data as is. However, when using the interframe coding method, in order to decode the image data of each frame, the image data of the frame immediately before that frame is required, so special playback such as high-speed reverse playback or high-speed playback is required. It is also difficult to selectively reproduce desired screens. Therefore, various encoding methods have been proposed that alternately use intraframe encoding and interframe encoding.
RM (R
reference model) has been proposed. This RM performs intra-frame encoding, performs motion compensation during inter-frame encoding to reduce the difference from the previous frame, and then performs discrete cosine transformation on the difference data, and converts the result to It is something that is encoded. According to this method, the image data of each frame subjected to intra-frame encoding can be decoded using only the code string corresponding to that frame, thus making it possible to perform the special reproduction described above. Also, since it is used in conjunction with interframe coding,
The overall number of required bits can be kept small. Such encoding methods: 1. Applications where there is a priori knowledge of the content of the moving image and real-time encoding is not required, such as CI) - ROM
It is widely regarded as an encoding method used when recording moving images on DVDs, etc. ``Problem to be solved by the invention'' However, as mentioned above, even when interframe coding and intraframe coding are used, frames that undergo intraframe coding still require a large number of bits. Therefore, there is a drawback that the number of bits becomes enormous when there are many changes in the screen. Furthermore, if the number of bits in a code string obtained by intraframe encoding increases, there is a risk that the transmission bit rate to the recording system will exceed the limit. In this case, in order to prevent transmission to the recording system from becoming impossible, a method is adopted in which the threshold value when quantizing the code string to be recorded is controlled in accordance with the data occupancy rate of the transmission path leading to the recording system. However, in this case, the code string obtained by intraframe encoding inevitably has a large amount of data, so a large quantization threshold (/f) is applied, and the quantization noise of that Freno is transmitted to other frames. For this reason, there has been a problem in that image quality deteriorates periodically during image reproduction.This invention was made in view of two consecutive circumstances. It is an object of the present invention to provide an encoding method for image data that can be efficiently encoded without any problems, and that can also perform special reproduction. For a desired frame, the pixel data constituting the frame is divided into a plurality of blocks, and for each block, the average value of each pixel data belonging to the block is encoded, and each pixel data and the average value are encoded. The method is characterized in that the difference between the pixel data of the frame that follows the desired frame is encoded, and the difference between the pixel data and the pixel data of the already encoded frame is encoded. Pixel data is divided into multiple blocks, and for each block (every 7 blocks), each pixel data belonging to the block is averaged and encoded, and the difference between each pixel data and the averaged pixel data is encoded. Embodiment: An embodiment of the present invention will be described below with reference to the drawings. (Explanation of encoding method) Below, The main features of the image data encoding method according to an embodiment of the present invention will be listed and explained. (1) Combination of intra-frame encoding and inter-frame encoding.
As shown in frames FR, O, PR11°, etc. shown in the figure, the first frame of the scene is subjected to intra-frame encoding, and the subsequent frames are subjected to inter-frame encoding. (2) Intraframe encoding method Frames such as frames P R, , ), F r (go, . Believe each block]
The elements are averaged to obtain the intra-block signal average value, and the intra-block signal average value of each macroblock is encoded to create reference data FUN. Also, find the difference between each signal element belonging to the frame and the intra-block signal average value corresponding to the macroblock to which the signal element belongs, and encode these differences to create intra-frame difference data IDEL. Here, the reference data FUN is composed of a small number of code strings corresponding to the number of macroblocks, and the intra-frame difference data can reduce the number of pits assigned to codes corresponding to one pixel. Encoding for a frame can be performed with a small number of required bits. By using only the reference data FUN and the intra-frame difference data IDEL obtained in this manner, the image of the frame can be reproduced. Therefore, from now on, the first frame P of scenes SC1, SC2, ...
RIO+ F Re.・is called a completely independent playback frame. (3) Application of semi-independent playback frames Select a desired frame in the scene (for example, in scene Scl)
1ro frame F II, 1. For F Ran, ·), the difference between each pixel data in the frame and the average signal value within each block applied to the reference data FUN created in the scene is encoded, and independent inter-frame difference data rεI) IΣ1, Create. A frame to which such an encoding method is applied will hereinafter be referred to as a semi-independent reproduction frame. The image data of a semi-independent reproduction frame is the difference data between independent frames of that frame E D
It can be decoded using E L and reference data FUN of completely independent frames in the scene. (Embodiment of Encoding Apparatus) FIG. 2 is a block diagram showing a configuration example of an encoding apparatus for implementing the above-mentioned encoding. In the figure, l is a subtracter, 2 is a discrete cosine transform circuit, 3 is a quantization circuit, and 4
5 is an inverse quantization circuit, 5 is a discrete cosine inverse transform circuit, 6 is an adder, and FM a and FMb are frame memories, respectively. Here, frame memories FMa and FMb are
Either one is selected by switch SW1, image data is read from the selected frame memory, and U
The signal is input to the subtracter 1 via the J-bus filter 16. 7 is a complete German Y7.7 that performs intraframe encoding. an independent frame extraction circuit that extracts image data of the reproduced frame;
8 is a vector quantization circuit that performs vector quantization of image data of independent frames; 9 is an inverse vector quantization circuit that restores image data based on the output vector obtained by vector quantization and various parameters; 10 is motion compensation The circuit 15 is an intra-cluster average value calculation circuit. Further, 11 and 12 are variable-length encoding circuits, 13 is a run-length encoding circuit, and 14 is a multiplexing circuit. Each output data of variable length encoding circuits II, 12, run length encoding circuit 13 and vector quantization circuit 8 is organized in a predetermined order by multiplexing circuit 14, and is stored in buffer B.
It is output via UFO and recorded on storage media such as CD-ROM. The operation of this encoding device will be explained below. This encoding device converts the color of each pixel of the screen into a luminance signal (hereinafter referred to as Y
signal) and two types of color difference signals (hereinafter referred to as U signal and V signal)
handles YUv image signals expressed by In addition, the image signal is organized in the CIF format used in image communication such as telegraph conferences,
This encoding device is manually inputted. FIG. 3(a) shows color signal matrices [Y], [U], and [V] consisting of Y, LJ, and V signals for one screen input every l frame period. As shown in this figure,
Matrix [Y] has 2881 rows and 356 columns, and matrices [U] and [V] both have 14 rows.
4. The number of columns is 178. In the case of CrF format,
Each element of each color signal matrix [Y], [U], and [V] for one screen is composed of 18X22=396 macroblocks [Y M kl (k=1 to 396), [U
Mk] (k = 1 to 396), [V Mk] (k-
1 to 396) and sequentially transmitted in time-division units in macroblock units. Here, Mac[1buUl-/ri[
YMk is 16x16=256 Y elements or ranals. Also, -/ 'y tr B L1 So7 r (J
Mkio, 1; and [VMk] are each 8x8=64
(1) Consists of LJ signal element, ■signal element. Fully independent reproduction frame encoding (intraframe encoding)
〉 When a completely independent playback frame is input, the macrob of that frame [1kl-Y Mkl, [tJ Mk
l, [VMk] are taken in by the independent frame extraction circuit 7, inputted to the vector quantization circuit 8, and vector quantized according to the procedure described below.

[Calculation of intra-block color signal average value]

First, according to the following formulas (1) to (3), the Y signal element Yij, the U signal element Llij, and the ■signal element Vij belonging to the macroblocks [YMk, [UMk], and [v Mk] are determined.
are averaged, and the intra-block color signal average values of Y, U, and V colors Y E (k), U E (k), V E (
k) are determined respectively.

【Normalization】

Next, the normalization calculations shown in equations (4) to (6) below are performed, and each element Y of the normalized color vector whose absolute value is "1" or less
S (k), U S (k), and V S (k) are obtained. i:1 j:1 i-1j=1 i:l j-■ Furthermore, the above three intra-block color signal average values Y!・〕(
k), LJ E (k), V E (k) average value X[
shi(k), 1; and bucket-shaped deviation 5E(k) are determined. Note that the average value xt-](k) is hereinafter referred to as the three-color average value It is called (k).

[Clustering]

Next, the normalized color vector (Y
Clustering of S (k), U S (k), and V S (k)l is performed. That is, as shown in FIG. 3(C), the YtJV color vector space is divided into N parts (m is an integer) so that each region has no coexisting part, and the normalized color vectors (Y S (k), U S (k), V S (k
)] are these divided areas C3 to Cm.
It is required to determine whether it belongs to this category. That is, the vector (Y
D +, U D 1. v D , ) ~ (Y I)
m, U D m, VI) s+) are stored in advance as a codebook [CB] as shown in FIG. 3(d), and from these, the normalized color vector (Y S k), U S (k), V S (k)
) is the normalized color vector (Y S (k), U S (k), V S (
k)) is selected as the output vector that approximates. Third
As shown in Figure (c), the normalized color vector (Y S (k
), LJ S (k), V S (k)) are area C
1, vector (YDlLJDVD D,
) is selected as the output vector. The vector quantization process described above is applied to all macro tl blocks [YMk] (k = 1 to 396), (:UMk] (k
= 1 to 396), [V Mk] (k = 1 to 396
) is executed. As a result, as shown in FIG. 3(e), an index matrix IN whose elements are the indexes of the output vectors corresponding to each macroblock is created.
J, a three-color average value matrix [XE] whose elements are the three-color average values X E (k) of each macroblock, and a standard deviation matrix [SE] whose elements are the standard deviation S E (k) of each macroblock? can get. FIGS. 4(a) to 4(c) show the results of trying the vector quantization described above for a certain screen. Here, FIG. 4(a) shows the index matrix [
N], Figure 4(b) is for 8 divisions, Figure 4(c) is for 1 division.
The cases of 6 divisions are each shown. Each element of the matrix [N] obtained in this way is input to the run-length encoding circuit 13, and for each element string of l series in which each element has the same value and continues, the common element value of the valley element and its continuation. The number of times is converted into a run-length code in combination with the table 4' value, and the code is input to the multiplexing circuit 14. Fourth
As shown in FIGS. (a) to (c), when an image is vector quantized, the same output vector often appears consecutively. Therefore, by performing run-length encoding on the vector index, the number of required bits can be significantly reduced. In addition, each element of the matrix [XE] and each element of the matrix [SE] are calculated by the intra-cluster average value calculation circuit 1.
The average value of the macroblock group [TI
:], [rT] are calculated. Matrix [XE],
Each element of [SE] is 396 per completely independent frame, but by calculating the intra-cluster average value in this way, it is compressed to the number of data corresponding to the number of indexes m (see Figure 3). d)). Then, the intra-cluster average values [1] and [-] are manually input to the multiplexing circuit 14. Note that the same codebook [CB] can be used if the video images to be recorded are of the same scene.
Multiplexing circuit for scene switching FRJ! Sent to 4.

[Inverse vector quantization]

The above matrix [N
], the codebook [CB] and the average value obtained by calculating the intra-cluster average value [5n'Z ], [n 1
is manually input to the inverse vector quantization circuit 9, and the following equations (7) to (9) are calculated for each macroblock number, and the intra-block color signal average value Y F (k), UI''(
k), V F (k) is decoded. Y F (k)−〇 B Y (N (k)) n (
N (k)) + T1(N (k)) ・
...(7) U P (k) -CB U (N (
k)) Chi E (N (k)) + 'TV (N (k)
) ・・・・・・(8) V F (k)−
CB V (N (k)) n (N (k))
N (k) f...(9) However,
In the above equations (7) to (9), N (k) is the index of the output vector of the macroblock. Also, CB Y (N (k)) is the codebook [CB]
Y signal element sequence C13Y (n) (n-1-n+
) means the element corresponding to index N (k). Similarly, CBtJ (N (k)), CT3
V (N (k) i also means the signal element corresponding to the index N (k) in each signal element sequence in the = r book [CB]. Also, n (N (k)) and T'T', (N (k)) are the average values corresponding to the index N (k) obtained by calculating the intra-cluster average value. Then, the intra-block color signal average value YF (k) of each macroblock is ), U P (k), V F
(k) is written to the frame memory FMb.

[Encoding of intra-frame difference data]

Further, during the intraframe encoding period, the switch SW is switched to the frame memory PMb side, and the processing described below is performed. The data stored in the frame memory FMb is input to the subtracter 1 via the low-pass filter 16. Note that this low-pass filter I6 is inserted for smoothing quantization errors. And a subtractor! input image data and frame memory FM
Difference data between the image data read from the image data b and input through the low-pass filter 16 is calculated. In detail, as shown in FIG. 3(g), macrob t'!
For each block number, the Y signal element Y Mk (i, D (i = 1 to 16 +j-1 to +6
) and the intra-block color signal average value Y F (k) in the frame memory FMb, and the Y signal difference matrix [Y A k] = Y A k (i, D (i
= I~! 6j-1 to 16) are obtained. Through similar processing, the U signal difference matrix [UA kl and the V signal difference matrix [VA kl] are obtained.Then, the discrete cosine transform circuit 2 calculates each color signal difference matrix [YAk ]
, [UAk] and [VAk] are subjected to discrete cosine transformation, and the coefficient matrix [CY kl, [c u kl, [cVk]
are obtained respectively. And these coefficient matrices [CY kl, [CLJkl
and [Each element of CV kl is quantized by the quantizer 3, and the resulting quantized value is sent to the variable encoding circuit 1.
1 is converted into a variable length code such as a Huffman code, and input to the multiplexing circuit 14. Here, the quantization threshold q in the quantizer 3 is controlled according to the free capacity of the buffer γ1 (LIFO, and a small value is set when the free capacity of the buffer BUFO is large, and a large value is set when the free capacity is small. In addition, the quantized value output from the quantizer 3 is decoded by the inverse quantization circuit 4 based on the same threshold q used for quantization.Here, each element of the coefficient matrix is , a deviation corresponding to the quantization error of the quantizer 3 occurs between the 1iiT input to the quantizer 3 and the output from the inverse quantizer 4. The code output from the inverse quantizer 4 are grouped in macroblock units, and the coefficient matrix of the spatial frequency component of the difference matrix of the Y, U, and V signals is restored.This coefficient matrix is then subjected to discrete cosine inverse transform by the discrete cosine inverse transform circuit 5. and the difference matrices of Y, U, and V signals [YB k], [U B k], [V B
k] is obtained. Then, by the adder 6, the following formulas (10) to (I2)
As shown in , the difference matrix [,y B k],
[LJBk], [VBk] and the intra-block color signal average value Y F (k) stored in the frame memory FMb,
U F (k) and v rx (k) are each added,
The color signal matrix [Y T(kl
, [U Hkl[V Hk] are written to the frame memory FMa. [Y Hk] = [Y B k] + y F (k)・
...(10) [U I-1k] - [U B k] +
u F (k)...(1ri [V Hk] - [V
Bk]+V F (k)---(+2) In the above manner, encoding of completely independent reproduction frames is completed. Then, each output data of the variable length encoding circuits 11 and 12 and the run length encoding circuit 13 and the average value [5n
'> ] and [■] are organized in a predetermined order by the multiplexing circuit 14 and recorded on a storage medium such as a CD-ROM. Furthermore, each element of the codebook [CB] is multiplexed with each matrix element, transmitted, and recorded on the media. Interframe encoding> Now, when performing interframe encoding, switch SW,
is switched to the frame memory FMa side.

[Motion compensation]

When performing interframe encoding, motion compensation is performed to keep the amount of data of the difference from the previous frame 11 small. That is, one is selected one after another from a group of motion compensation vectors that have been transformed in advance, and according to this motion compensation vector,
Each element of the image data stored in the frameau memory ['M a is translated in parallel. Then, the image data that has been translated in parallel using the motion vector is compared with the human image data for this encoding device, and a motion compensation vector [M OV ] is determined that provides the strongest correlation between both image data. is selected. Each element of the motion compensation vector [M OV ] determined in this manner is converted into a variable length code such as a Huffman code by the variable length encoding circuit 12, and is then converted into a variable length code such as a Huffman code by the variable length encoding circuit 12. 4 is input.

[Differential encoding]

Then, the subtracter l subtracts the macroblocks [Y M k] and [u M k of each signal constituting the input image data.
Difference between l[VMk] and the motion-compensated Y, Lj, V signal matrices [YHk], [UHkl, [VHk] stored in the frame memory FMa -/ t') -/ Kus [Y A k] [U A
k] and [V A k] are calculated. Thereafter, as in the case of intra-frame encoding described above, encoding is performed using discrete cosine transform, and the resulting code is multiplexed with other information by the multiplexing circuit I4, and is sent to the buffer B U
It is transmitted to the recording system via FO. Note that during this inter-frame encoding period, the independent frame extraction circuit 7, the flat quantization circuit 8, the inverse vector quantization circuit 9, and the intra-cluster average value calculation circuit 15 do not operate, and the matrix [N] , [n], [1m] and each element of the codebook [CB] are not output. <Encoding of semi-independent playback frames> During the encoding of semi-independent playback frames, the switch SW
I is switched to the frame memory FMb side. Then, each input macroblock [YM k], [t
J M k], [VM kl and frame memory F
The intra-block color signal average value Y[? of each macroblock in the completely independent reproduction frame stored in Mb? (k)U
The difference between F (k) and V F (k) is calculated by the subtracter 1, and the difference data is encoded by discrete cosine transform as in the case of the double series, and sent to the multiplexing circuit I4 and the buffer BUFO. It is transmitted to the recording system via the After that, a completely independent playback frame is input [711,
Similar processing is performed when semi-independent playback frames are manually generated. (Embodiment of Decoding Apparatus) FIG. 5 is a block diagram showing the configuration of a decoding apparatus that decodes the code created by the encoding apparatus of FIG. 2. In the figure, BUFI is a buffer that sequentially takes in codes read from the media 1, 21 is a separation circuit, 22 and 23 are variable length code decoding circuits that decode variable length codes into fixed length codes, and 24 is a buffer that converts run-length codes into fixed-length codes. 25 is an inverse vector quantization circuit, 26 is an inverse quantization circuit, 27 is a discrete cosine inverse transformation circuit, and 28 is an adder. Further, FMc and FMd are frame memories, and any of the image data stored in these is selected by the switch S W t and sent to the adder 28 . Further, 29 is a motion compensation circuit, and 30 is a (7-Has filter). The operation of this decoding device will be explained below.
The separation circuit 2I determines the data classification of the code string. And the codebook [CI3
] and the average value for each index [51”
"H, ], [n ] are sent to the inverse vector quantization circuit 2. Also, the run-length code of the vector index N (k) of each macroblock is converted into a normal code string by the run-length code decoding circuit 24. Then, the calculations of equations (7) to (9) described above are executed, and the intra-block color signal average value Y F ( k), UP
(k), V I' (k) is restored and written to the frame memory l' M (l. Also, the variable length code created by the variable length encoding circuit I+ in FIG. DIF is separation circuit 2
1 and input to the variable length code decoding circuit 22. Then, after being converted into a fixed length code, the code value is returned to the code value before muntinization by the inverse quantization circuit 26. Here, the quantization code is decoded based on the threshold value q read from the medium. That is, in the 7'l arching device shown in FIG. 1, the f quantization circuit 3...
When changing the threshold value q, a code indicating the threshold value q is written on the medium, and during playback, the threshold value q is read from the medium and set in the reverse arithmetic circuit 26. 23 4 And each input difference code DIF, DIF. , the inverse operation of making the variable-length code fixed-length and making it into a prince is performed, and the coefficient matrix of the difference data described above [
CY k], [CU k], rCV k] are restored. Then, this coefficient matrix [CY k], [CU kl,
[cVk] is subjected to discrete cosine inverse transform by the discrete cosine inverse transform circuit 27, and the difference matrix of the Y, tJ, and V signals [Y [3k], [113k1.
[V 13 k] is obtained. At this time, the switch SW is switched to the Freno memory l''Md side.Then, the Freno memory FM
Intra-block color confidence/li average value Y Ii' (k) in d
, [IF (k), v F(k) is read out via the low-pass filter 30 and sent to the adder 28 as L, t, X
-Min Matri Socs [Y B kl, [u B k],
・[3kl are added to each corresponding one, and the color signal matrix l Y 1.1 kl , l UHk],
[VHk] is restored. Then, for each macroblock, the color signal matrix [Y
r-1kl, [011k2, [V 11 kl are restored, and images of completely independent reproduction frames are reproduced based on these. Also, the color signal matrices [YHk], [UT(k], and [VHk] of each macroblock are written to the frame memory FMc. In this way, completely independent reproduction frames do not require decoding of other frames, so It is possible to select and read only completely independent playback frames and perform special playback such as high-speed playback, high-speed reverse playback, and selective playback of a desired screen. <Decoding of interframe encoded code> In this case, the motion hli The variable length code of each element of the compensation vector [M OV ] is inputted to the variable length code decoding circuit 23 via the separation circuit 21, decoded into a fixed length code, and inputted to the motion compensation circuit 29. , the motion compensation circuit 29 calculates the color signal matrix [Ytlkl[LJ Hk], [V Ll
k] is translated in accordance with the motion compensation vector [MOV]. In addition, the difference code [) corresponding to the difference with nQ frame 11
IF passes through the separation circuit 2I, =S variable length code decoding circuit 22
is input. Then, the same process as in the case of decoding the completely independent reproduction frame described in -1- is performed, and the discrete cosine inverse transform circuit 27 outputs difference matrices [Y B k], [U B k] corresponding to the difference from the previous frame. ] [V B
kl is obtained. At this time, the switch SW t has been switched to the frame memory FMc side. Then, motion-compensated color signal matrices [Y Hk], [U Hk],
[Vl-1k] is read out via the low-pass filter 30, and the adder 28 adds the difference matrix [YI3
kl. [UBk1. [V
lk1. mouth J [-1k], [V Hk] is obtained. Then, this new color signal matrix [Y I-1
k], [U I-1kl, 1VI(k], and a new color signal matrix [Y
Hk], [U Hk], and [V It kl are written to the frame memory F M c. Decoding of quasi-independent playback frames> In this case, the switch S W is the frame memory r>M
It can be switched to the d side. Frame memory ri'Md1.
: is the intra-block color signal average value YF(k) of each macroblock of a completely independent reproduction frame in the scene,
U F (k) and V F (k) are stored. Then, the intra-block color signal average value YF of each macroblock
(k), [J F (k), and the difference code DIF between V F (k) is input to the variable length code decoding circuit 22 via the separation circuit 21 . Then, the same process as in the case of decoding the completely independent reproduction frame described above is performed, and a new color signal matrix [YT-1k11. Mouth J It
kl , [V 11 kl are restored, and an image is reproduced using them, and a new color signal 7 matrix l'i'Hk1 .口JHk'], [VHk] are written to the frame memory InMe. (Modified example) Reference data F for completely independent CF cow frame in the present invention
Since tJ N is extremely compressed, as shown in Fig. 6, the frames F It v., F It treated as semi-German X''f, +lr raw frames in Fig. 1 are
,,. , . . . etc. may be treated as completely independent playback frames, and the independent playback frames may be eliminated. "Effects of the Invention" As explained above, according to the present invention, for a desired frame, pixel data constituting the frame is divided into a plurality of blocks, and for each block, each pixel data belonging to the block is At the same time, the difference between each pixel data and the average value is encoded, and for the frame following the desired frame, the difference between the pixel data of the already encoded frame is encoded. As a result, encoding can be performed efficiently, but -1
- by reproducing the code of the desired frame;
This has the effect of allowing special playback. In addition, the number of bits required for encoding a desired frame is small (the number of hits between other frames), so even during normal playback, it is possible to avoid transmission bit rate constraints, etc. The bit precision of the code is not sacrificed. Therefore, it is possible to perform encoding without causing periodic image deterioration.

[Brief explanation of drawings]

FIG. 1 is a diagram illustrating an image data encoding method according to an embodiment of the present invention, FIG. 2 is a block diagram showing the configuration of an encoding device according to the embodiment, and FIG. 3 is a diagram illustrating the encoding method of FIG. 2. FIG. 4 is a diagram showing the trial results of vector quantization in the same embodiment. FIG. 5 is a block diagram showing the configuration of the decoding device according to the same embodiment. The figure is a diagram illustrating a method of encoding image data according to a modified example of the present invention. F Rlo, P It n n...Completely independent recycled freight, FLJN...Jl (semi-data, l l)
l-]1.・Freight 1, internal difference data.

Claims (1)

[Claims] For a desired frame, the pixel data constituting the frame is divided into a plurality of blocks, and for each block, the average value of each pixel data belonging to the block is encoded, and the average value of each pixel data belonging to the block is encoded. A method for encoding image data, comprising: encoding a difference between data and the average value; and for a frame following the desired frame, encoding a difference between pixel data of an already encoded frame.