JPH0884339A - Image re-compression method and image re-compression device - Google Patents

Abstract

PURPOSE: To provide a method and device expanding bit stream of an image subject to compression coding to compress the image again without reproducing the image and to prevent propagation of error due to re-compression. CONSTITUTION: Coded image data are received from a terminal 100 and a variable length decoder 110 decodes a 1st quantization parameter and a 1st quantized coefficient. An inverse quantization device 130 uses the 1st quantization parameter to apply inverse quantization to a 1st quantized coefficient to generate a 1st inverse quantization coefficient. An adder 210 subtracts a correction coefficient generated based on preceding image data stored in a difference coefficient memory 180 from the 1st inverse quantization coefficient and gives the difference to a quantization device 140, in which the data are quantized by a 2nd quantization parameter to generate a 2nd quantized coefficient. The 2nd quantization parameter and the 2nd quantized coefficient are given to a variable length coder 150, in which they are converted into a variable length code, which is fed to a multiplexer 160.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image recompression method and an image recompression device for recompressing compressed and encoded digital image data for transmission or recording.

[0002]

2. Description of the Related Art In order to efficiently transmit and record digital moving images, compression coding is necessary. There are intra-frame coding and inter-frame predictive coding as a moving image compression method. In intra-frame coding, compression processing of one image is completed in one frame, whereas in inter-frame coding, motion-compensated prediction is performed from adjacent frames to realize high-efficiency compression. Depending on the application, compression may be performed by combining intraframe coding and interframe coding. In some cases, a series of moving images are all intra-frame encoded, or in some cases, the first image is intra-frame encoded and all the remaining images are inter-frame encoded. In some cases, intra-frame coding is periodically performed, and inter-frame predictive coding is performed on an image located between two intra-frame coded images.

Intraframe coding removes redundancy in space. As means for that purpose, an orthogonal transform method represented by discrete cosine transform (hereinafter referred to as DCT), a wavelet transform for dividing into frequency bands, or a subband method is used. The orthogonally transformed coefficients and the coefficients divided into frequency bands are quantized to a desired transmission or storage amount and variable length coded. The quantization step must also be coded so that it can be decoded on the playback side. In addition, D
When performing CT, the image is divided into blocks of appropriate size and then converted.

On the other hand, interframe predictive coding removes redundancy in the time direction. An image of one frame is divided into a plurality of adjacent blocks, a past or future frame is referred to for each block, and a motion vector is obtained under a predetermined evaluation function. Using the obtained motion vector, the reference block at the offset position is used as the prediction signal. The difference between the target block and this prediction signal block (hereinafter referred to as a prediction error) is calculated, and the intra-frame redundancy is further removed by the intra-frame coding method described above.
A reproduced image is often used as a prediction signal, and an image once encoded must be decoded and reproduced. In addition to the past and future prediction signals, the prediction signal may be created by an average of the motion-compensated past and future signals or a weighted average. Further, there may be a block that is encoded in the same manner as the intraframe encoding without obtaining the differential signal.
That is, there are a plurality of coding modes in an inter-frame coded image. Therefore, in addition to the differential signal, the motion vector and the information on the coding mode must be coded. In the case of wavelets and sub-bands, the redundancy in the time direction may be removed by dividing into frequency bands and then dividing them into blocks for motion compensation.

As described above, when compressing and coding a digital moving image, quantized transform coefficients (or coefficients divided into frequency bands), quantization widths, and coding methods (intraframe / interframe) are used. ), Information such as the mode of the coding block and the motion vector is coded. This series of encoded data streams is called a bitstream. On the reproducing side, this bit stream is read and decoded to decompress and reproduce the image.

[0006]

On the other hand, depending on the application,
The compression encoded image may be recompressed to a lower bit rate (ie, higher compression rate). For example, at a broadcasting station or a movie production company, an original image is compression-encoded into a high-quality image at a high bit rate and stored, and if necessary (for example, bandwidth of a transmission path or capacity limitation of a recording medium). Recompress to a lower bitrate. There is also a demand to record more images than standard recording on one recording medium (for example, VTR tape). For example,
Like the current home analog VTR, standard and long-time recording modes are provided to reduce the tape speed by, for example, 3 times, thereby enabling recording for 3 times the time and realizing long-time recording. In the case of a bitstream, if the tape is recorded at a slower speed, some of the data in the bitstream cannot be recorded, so the image must be recompressed.

There are the following methods for recompressing a compression-coded image. One of them is a transcoding method. That is, both the decoding device and the encoding device are prepared, the bit stream of the compression-encoded image is decoded by the decoding device and the image is reproduced, and then the reproduced image is reproduced by the encoding device up to a desired bit rate. This is a compression coding method. This method requires both a decoding device and a coding device, which is expensive, and doubles the delay because decoding and then coding are performed again. It is not practical as a home VTR.

There is also a method of recompressing an image without completely decoding and reproducing it. An image compressed and encoded by the interframe prediction DCT encoding method described in the related art will be described as an example with reference to FIGS. 8 and 9. FIG. 8 shows a flowchart of a conventional image recompression method. First, a bitstream of a compression-encoded image is input (step 1
0). Quantized DCT from bitstream
The coefficient a [i] of and the first quantization width q [i] are extracted. Where i = 1, ..., N, where N is the DCT of the target block
The number of coefficients. The value of N depends on the target block. Next, the coefficient a [i] is multiplied by q [i] for inverse quantization (step 12). The inverse quantized coefficient b [i] obtained by inverse quantization is divided by the second quantization width q _N [i] to generate the requantized coefficient c [i] (step 14) and output. (Step 16). Normally, q so that the recompressed image bitstream has a smaller amount of data than the input bitstream.
Control _N [i].

FIG. 9 shows a conventional image recompression device (COMP).
A block diagram of 300 is shown. A bitstream of a compression-encoded image is input from the input terminal 100. The variable length decoder (VLD) 110 analyzes the bit stream and determines the quantization width q [i] and the quantized DCT coefficient a [i].
Select and decrypt. The result is sent to the inverse quantizer (IQ) 130 via the line 1010. Data other than q [i] and a [i] are sent to the multiplexer 160 via the line 1000. In the inverse quantizer 130, step 12 in FIG.
The inverse quantization operation is performed as described above to generate b [i], which is sent to the quantizer (QN) 140 via the line 1020. In the quantizer 140, q _N [i] is set to b as in step 14 of FIG.
Quantize [i] to generate c [i]. c [i] and q _N [i] are sent to the variable length coder (VLC) 150 via the line 1030, converted into a variable length code and sent to the multiplexer (MUX) 160 via the line 1040. The multiplexer 160 multiplexes the data sent via the lines 1000 and 1040 and outputs the multiplexed data. In addition, in order to compress to a desired bit rate, the rate controller 120
While counting the bits of the output of the variable length encoder 150, q _N
Determine [i].

When recompressed by the above method and apparatus, it is not necessary to reproduce and encode an image, so that the encoding apparatus can be simplified and the delay time can be shortened. It is effective if the compression rate at the time of recompression is low. However, when the recompression is performed at a high compression rate as compared with the interframe prediction method, the image quality deteriorates.
In the inter-frame prediction method, as described in the related art, the past image is used as the prediction signal in the compression order, but since the past image is also recompressed, the prediction signal changes and more Will include the quantization noise (due to requantization). Quantization noise introduced by recompression propagates one after another, causing deterioration in image quality.

There is also a method of recompressing by discarding a predetermined number of coefficients (mainly high frequency components) from the DCT coefficients instead of requantization. In this case as well, error propagation due to recompression similarly occurs.

[0012]

In order to solve the above-mentioned problems, the present invention corrects the prediction error of the current image with the error signal generated by recompressing the past image as follows. Then recompress. The image data quantized and compressed with the first quantization parameter is inversely quantized with the first quantization parameter to generate first inverse quantized image data.
From the first dequantized image data, the correction signal generated by the past image data is subtracted and then quantized with the second quantization parameter and output, and also the second quantization parameter is dequantized and the second The inverse quantized image data of is generated. A difference signal is generated from the second dequantized image data and the first dequantized image data, and this difference signal is used as a correction signal for the next image data. Preferably, the correction signal generated by the past image data is added to the second dequantized image data, and then the difference signal is generated from the first dequantized image data.

Further, the correction coefficient generated by the past image data is subtracted from the first conversion coefficient block included in the image data that has been orthogonally transformed and compressed to generate the second transform coefficient block. A predetermined number of coefficients are extracted from the second transform coefficient block, a third transform coefficient block is formed and output, and a difference coefficient is extracted from the third transform coefficient block and the first transform coefficient block. Generate a block. This difference coefficient block is used as a correction coefficient for the next image data. Preferably, the correction coefficient generated by the past image data is added to the third transform coefficient block, and then the difference coefficient block is generated from the first transform coefficient block.

Further, the following recompression device prevents the propagation of error due to requantization. Variable length decoder, first dequantizer, first adder, quantizer, variable length encoder, multiplexer, second dequantizer, second and A third adder and a difference coefficient memory are provided. Inputs encoded image data, decodes the first quantization parameter and the first quantized coefficient by the variable length decoder, sends it to the first dequantizer, and sends other data to the multiplexer. . The first dequantizer dequantizes the first quantized coefficient with the first quantization parameter to generate a first dequantized coefficient. The first adder subtracts the correction coefficient generated by the past image data stored in the difference coefficient memory from the first dequantized coefficient, and then inputs the subtracted correction coefficient to the quantizer. So quantize with the second quantization parameter,
Generate a second quantized coefficient. The second quantization parameter and the second quantized coefficient are converted into a variable length code by a variable length encoder, sent to a multiplexer, multiplexed with other data, and output. On the other hand, the second dequantizer dequantizes the second quantized coefficient with the second quantization parameter to generate the second dequantized coefficient. The second adder adds the correction coefficient generated by the past image data to the second dequantized coefficient, and then the third adder generates a difference coefficient from the first dequantized coefficient. , The difference coefficient memory. This difference coefficient is used as a correction coefficient for the next image data.

Further, the following recompression device prevents the propagation of an error due to requantization. Variable length decoder, first dequantizer, first inverse orthogonal transformer, first adder, orthogonal transformer, quantizer, variable length encoder, multiplexer , A second inverse quantizer, a second inverse orthogonal transformer,
Second and third adders and a differential signal memory are provided. Inputs encoded image data, decodes the first quantization parameter and the first quantized coefficient with a variable length decoder, sends to the first dequantizer, and sends other data to the multiplexer . The first inverse quantizer inversely quantizes the first quantized coefficient with the first quantization parameter to generate a first inverse quantized coefficient. The first inverse orthogonal transformer transforms the first inverse quantized coefficient into a first spatial signal. The first adder subtracts the correction signal generated by the past image data stored in the difference signal memory from the first spatial signal. The result is orthogonally transformed by an orthogonal transformer. The obtained coefficient is input to the quantizer and quantized with the second quantization parameter to generate the second quantized coefficient. The second quantization parameter and the second quantized coefficient are input to a variable length encoder, converted into a variable length code, sent to a multiplexer, multiplexed with other data, and output. On the other hand, the second dequantizer dequantizes the second quantized coefficient with the second quantization parameter to generate the second dequantized coefficient. The second inverse orthogonal transformer transforms the second inverse quantized coefficient into a second spatial signal. Next, the second adder adds the correction signal generated by the past image data to the second spatial signal, and then the third adder generates a difference signal from the first spatial signal. , The difference signal memory. This difference signal is used as a correction signal for the next image data.

Further, the following recompression device prevents the propagation of an error caused by truncating the transform coefficient. A variable length decoder, a first adder, a coefficient selector, a variable length encoder, a multiplexer, second and third adders, and a difference coefficient memory are provided. The encoded image data is input, the variable length decoder decodes the first transform coefficient block, and the decoded block is sent to the first adder. Send other data to the multiplexer. The first adder subtracts the correction coefficient generated by the past image data stored in the difference coefficient memory from the first transform coefficient block to generate the second transform coefficient block. Next, the coefficient selector extracts a predetermined number of coefficients from the second transform coefficient block,
Form a third transform coefficient block. The third transform coefficient block is converted into a variable length code by a variable length encoder, sent to a multiplexer, multiplexed with other data, output, and sent to a second adder. The second adder adds the correction coefficient generated by the past image data to the third transform coefficient block, and then the third adder generates the difference coefficient from the first transform coefficient block. And store it in the difference coefficient memory. This difference coefficient is used as a correction coefficient for the next image data.

[0017]

By using the image recompression method and the image recompression apparatus according to the present invention, the input bitstream can be recompressed to another bit rate without the need to completely solve the bitstream to reproduce the image. Therefore, there is almost no delay. Further, since the prediction error of the current image is corrected by the error signal generated by recompressing the past image and then the image is recompressed, the error propagation due to the recompression can be prevented and the deterioration of the image quality is reduced. be able to.

[0018]

Embodiments of the present invention will be specifically described below with reference to the drawings. In these embodiments, a bitstream obtained by compression coding with the coding method of interframe prediction DCT described in the prior art is used as an example, but the recompression method and the recompression apparatus of the present invention use this method. However, other compression encoding methods using inter-frame prediction (for example, subband or wavelet conversion),
It can be similarly applied to a compression coding method using pixel or inter-data prediction (for example, differential pulse coding, abbreviated as DPCM).

(First Embodiment) A first embodiment of the present invention will be described with reference to FIG. FIG. 1 is a flow chart for explaining the first embodiment of the image recompression method of the present invention. First, in step 40, the bit stream of the image compressed and encoded is input. The quantized DCT coefficient a [i] and the first quantization width q [i] are extracted from the bitstream. Here, i = 1, ..., N, and N is the number of DCT coefficients of the target block. The value of N depends on the target block. As explained in the prior art, a [i]
Is a coefficient obtained by DCT of the original image or the prediction error and quantized by q [i]. The first quantization width is q
It may be [i] = q [j]. Next, the coefficient a [i] is multiplied by q [i] for inverse quantization (step 42). Then, the correction coefficient z [i] is subtracted from the inverse quantization coefficient b [i] obtained by inverse quantization (step 43). The correction coefficient z [i] is a recompression error generated by recompressing the past image data (the method of obtaining it is performed in step 48). The coefficient b ′ [i] corrected in this way is used for the second quantization width q in step 44.
Quantize with _N [i]. For i ≠ j, q _N [i] = q _N [j] may be satisfied. Normally, q _N [i] is controlled so that the recompressed image bitstream has a smaller amount of data than the input image bitstream. The coefficients c [i] and q _N [i] thus obtained are output in step 50.

Next, how to obtain the correction coefficient will be described.
This correction coefficient is a quantization error caused by quantization with the second quantization width q _N [i]. In order to obtain this error, first, in step 45, inverse quantization is performed. That is, the inverse quantized coefficient d [i] is generated by multiplying c [i] by q _N [i]. Since the correction coefficient z [i] is subtracted from b [i] in step 43, z [i] is added to d [i] in step 46 to generate e [i]. e [i] is a coefficient obtained by requantizing and reconstructing b [i]. e [i]
And the difference between b [i] and q _N [i] gives the quantization error caused by requantization. That is z '[i] obtained in step 48. Since z '[i] becomes a correction signal for recompression of the next image, it is temporarily stored (step 5).
0). The above process is repeated for all blocks of the current image, and the correction coefficient z '[i] corresponding to each block
Is stored. Before recompressing the next image, z '[i] of each block is substituted into z [i] before processing. Since the intra-frame coded block is coded without referring to the past image, steps 43 and 46 are not performed. Further, although the above-described processing is performed in the DCT domain, the DCT coefficient may be inversely transformed and performed in the spatial domain.

It is not necessary to perform the inverse quantization as in step 42. In this case, step 42 is b [i] = a
It becomes [i], and q [i] is multiplied by q _N [i]. Q _{N in} step 50
Output q [i] instead of [i].

(Second Embodiment) A second embodiment of the present invention will be described with reference to FIG. FIG. 2 is a flow chart for explaining the second embodiment of the image recompression method of the present invention. First, in step 70, a bit stream of an image compressed and encoded is input. The quantized DCT coefficient a [i] is extracted from the bitstream. Where i = 1, ...,
In N, N is the number of DCT coefficients of the target block. The value of N depends on the target block. a [i] is a coefficient obtained by DCT-quantizing the original image or the prediction error.
Next, at step 72, the correction coefficient z [i] is subtracted from the coefficient a [i]. The correction coefficient z [i] is a recompression error generated by recompressing the past image data (the method of obtaining the error is performed in step 78). Coefficients corrected in this way
The K coefficients are left out of a '[i] and the other coefficients are discarded (step 74). K ≦ N. Therefore, the image is further compressed by truncating the coefficient.
The value of K differs depending on the block, and the bitstream of the image obtained by recompression is controlled so that the data amount is smaller than that of the bitstream of the input image. The coefficient u [i] thus obtained is output in step 80.

Next, how to obtain the correction coefficient will be described.
This correction coefficient is an error caused by truncating a '[i]. To obtain this error, first, step 76
Generate z [i] by adding z [i] to u [i]. w [i] is corrected a
It is the coefficient remaining after rounding down [i]. w [i] and a
If the difference from [i] is calculated, the error caused by the truncation can be obtained. That is z '[i] obtained in step 78. z '
[i] is the correction signal for recompression of the next image, so
It is temporarily stored (step 80). The above processing is repeated for all blocks of the current image, and the correction coefficient z '[i] corresponding to each block is stored. Z '[i] of each block is z [i] before recompression of the next image
Subsequent to and then processed. Since the intra-frame coded block is coded without referring to the past image, steps 72 and 76 are not performed.

(Third Embodiment) A third embodiment of the present invention will be described with reference to FIG. FIG. 3 is a block diagram of an image recompression device 300 for explaining the third embodiment of the present invention. Variable length decoder 110, rate controller 120, first dequantizer 130, first adder 210, quantizer 140,
It comprises a variable length encoder 150, a multiplexer 160, a second dequantizer 170, a second adder 220, a third adder 230 and a difference coefficient memory 180.

The encoded image bit stream is input from the terminal 100. The variable length decoder 110 extracts the quantization parameter q [i] and the quantized DCT coefficient a [i] from the bit stream, and decodes the code into a numerical value.
Where i = 1, ..., N, N is the DC of the target block
It is the number of T coefficients. The value of N depends on the target block. The output of the variable length decoder 110 is sent to the inverse quantizer 130 via line 2010. Other data is sent to multiplexer 160 via line 2000.
The inverse quantizer 130 dequantizes a [i] with q [i] to generate an inverse quantized coefficient b [i]. The b [i] is sent to the first adder 210 via the line 2020, where the correction coefficient z [i] generated from the past image data stored in the difference coefficient memory 180 from b [i]. Is subtracted to generate b '[i]. Quantizer 1 for b ′ [i] via line 2022
It is input to 40 and quantized with a new quantization parameter q _N [i] to generate a quantized coefficient c [i]. c [i] and q _N [i] are sent to the variable length encoder 150 via the line 2030, converted into a variable length code, and sent to the multiplexer 160. The multiplexer 160 multiplexes the codes of c [i] and q _N [i] sent via the line 2040, and other data sent via the line 2000, and outputs the multiplexed data. In this way the input bitstream is recompressed. Normally, the rate controller 120 is used to determine the quantization width q _N [i] so as to obtain a desired bit rate while counting the number of bits of the output of the variable length encoder. The inverse quantizer 130 may be omitted. That's the line 2020
Signal b [i] of is equal to a [i]. In this case, the result of multiplying the q _N [i] to q [i], in place of q _N [i], and sends to the multiplexer 160 via line 2040.

On the other hand, q _N [i] and c [i] are sent to the inverse quantizer 170 via the line 2032. Therefore, c [i] is inversely quantized by q _N [i] to generate an inverse quantized coefficient d [i]. In the second adder 220, the correction coefficient z [i] generated from the past image data is added to d [i]. The result is the third adder 2
30 and sent via line 2024 b
The difference coefficient z ′ [i] is generated from [i] and the difference coefficient memory 18
Store in 0. This difference coefficient is used as a correction coefficient for the next image data.

Next, a preferred embodiment of the difference coefficient memory 180 will be described. FIG. 4 shows a block diagram of the difference coefficient memory 180. Inverse orthogonal transformer 182, frame memory 18
4. The orthogonal transformer 186 is provided. The difference coefficient z ′ [i] input from the terminal 181 is a DCT domain coefficient.
First, it is inversely transformed by the inverse orthogonal transformer 182 to be returned to the spatial domain difference signal. The result is stored in the frame memory 184 and used when recompressing the next image.

Now, consider the case of recompressing the next image. As described in the related art, the prediction signal is obtained by motion compensation, and therefore motion compensation is required to obtain the spatial domain difference signal from the frame memory 184. This motion information is sent from the variable length decoder 110 via line 2080. Based on that, the difference signal of the target block is obtained from the frame memory 184 and the line 2094
To the orthogonal transformer 186. Therefore, the differential signal in the spatial domain is transformed into the transform domain to generate the differential coefficient.
This difference coefficient is added via line 2070 to adder 210.
And used as a correction coefficient. When performing motion compensation with an accuracy of half a pixel or more, the difference signal stored in the frame memory 184 is interpolated and up-sampled before motion compensation. It is also possible to store the coefficient z '[i] in the DCT area as it is in the frame memory 184 and perform motion compensation in the DCT area to generate a correction coefficient. In this case, the inverse orthogonal transformer 182 and the orthogonal transformer 1
And 86 are omitted.

In this embodiment, the interframe predictive coding has been described, but it may be the case of predictive coding (DPCM) between pixels or between other data. In this case, since the signal is not in the transform domain, the inverse orthogonal transformer 182 of FIG.
And the orthogonal transformer 186 are omitted.

(Fourth Embodiment) A fourth embodiment of the present invention will be described with reference to FIG. FIG. 5 is a block diagram of an image recompression device 300 for explaining the fourth embodiment of the present invention. Variable length decoder 110, rate controller 120, first inverse quantizer 130, first inverse orthogonal transformer 240, first adder 210, orthogonal transformer 250, quantizer 140, variable length code It includes a quantizer 150, a multiplexer 160, a second inverse quantizer 170, a second inverse orthogonal transformer 260, a second adder 220, a third adder 230 and a difference signal memory 270.

The encoded image bit stream is input from the terminal 100. The variable length decoder 110 extracts the quantization parameter q [i] and the quantized DCT coefficient a [i] from the bit stream, and decodes the code into a numerical value.
Where i = 1, ..., N, where N is the DCT of the target block
The number of coefficients. The value of N depends on the target block. The output of the variable length decoder 110 is sent to the inverse quantizer 130 via line 3010. Other data is sent to multiplexer 160 via line 3000. The inverse quantizer 130 dequantizes a [i] with q [i] to generate an inverse quantized coefficient b [i]. b [i] is sent to the first inverse orthogonal transformer 240 via line 3012. The first inverse orthogonal transformer 240 transforms the coefficient b [i] in the DCT domain into the spatial domain and transforms it into the first spatial signal s [i]. s [i] is line 30
Sent to the first adder 210 via 20 where s
The correction signal z [i] generated by the past image data stored in the difference signal memory 270 is subtracted from [i] to generate s' [i]. s' [i] is input to the orthogonal transformer 250 via the line 3022, and s' [i] is transformed into DCT domain coefficient b '[i]. b ′ [i] is input to the quantizer 140 via the line 3024, and the new quantization parameter q
Quantize with _N [i] to generate a quantized coefficient c [i]. c [i] and q _N
[i] and the variable length encoder 15 via line 3030
0, convert to variable length code, multiplexer 160
Send to In the multiplexer 160, the codes of c [i] and q _N [i] sent via the line 3040 and the line 3
The data is multiplexed with other data sent via 000 and output. In this way the input bitstream is recompressed. Usually, the rate controller 120 is used to determine the quantization width q _N [i] so as to obtain a desired bit rate while counting the number of bits of the output of the variable length encoder.

On the other hand, q _N [i] and c [i] are sent to the second dequantizer 170 via the line 3032. Where q _N [i] is c
[i] is inversely quantized to generate an inverse quantization coefficient d [i]. Next, d [i] is sent to the second inverse orthogonal transformer 260 to transform the DCT domain coefficient d [i] into the spatial domain signal s "[i]. In the second adder 220, s "[i] is added with the correction signal z [i] generated by past image data. The result is sent to the third adder 230 to generate a differential signal z ′ [i] from s [i] and store it in the differential signal memory 270. This difference signal is used as a correction signal for the next image data. If the next image is to be recompressed, the variable length decoder 110 to line 30
Based on the motion information sent via 80, the difference signal of the target block is fetched from the difference signal memory 270, sent to the first adder 210 via the line 3070, and used as a correction signal. In the case of motion compensation with an accuracy of half a pixel or more, the difference signal stored in the difference signal memory 270 is interpolated, up-sampled, and then motion compensated.

(Fifth Embodiment) A fifth embodiment of the present invention will be described with reference to FIG. FIG. 6 shows a block diagram of an image recompression device 300 for explaining a fifth embodiment of the present invention. Variable length decoder 110, rate controller 120, first adder 210, coefficient selector 280, variable length encoder 150,
It comprises a multiplexer 160, a second adder 220, a third adder 230, and a difference coefficient memory 180.

The encoded image bit stream is input from the terminal 100. The variable length decoder 110 extracts the quantized DCT coefficient a [i] from the bit stream and decodes it from a code into a numerical value. Where i = 1, ..., N
Where N is the number of DCT coefficients in the target block. The value of N depends on the target block. The output of the variable length decoder 110 is sent to the first adder 210 via line 4010. Other data is sent to multiplexer 160 via line 4000. First adder 2
In 10, the correction coefficient z [i] generated from the past image data stored in the difference coefficient memory 180 is converted from a [i].
Is subtracted to generate a '[i]. N a '[i], i = 1,
.., N is sent to the coefficient selector 280 via the line 4020. Therefore, K pieces are left out of N pieces a ′ [i], and the other coefficients are cut off to form a block made up of u [i]. However, K ≦ N. By truncating the coefficients in this way, the image is further compressed. Then, u [i] and u [i] are sent to the variable length encoder 150 via the line 4030, converted into a variable length code, and sent to the multiplexer 160. In the multiplexer 160,
The code of u [i] sent via line 4040,
It is multiplexed with other data sent via the line 4000 and output. The value of K differs depending on the block, and the normal rate controller 120 is used to determine the value of K so as to obtain a desired bit rate while counting the number of bits of the output of the variable length encoder.

On the other hand, u [i] is sent to the second adder 220 via the line 4032. Therefore, u [i] is added with the correction coefficient z [i] generated by past image data, and w
Generate [i]. Next, in the third adder 230, w [i]
And the difference coefficient z ′ [i] is generated from a [i] and stored in the difference coefficient memory 180. This difference coefficient is used as a correction coefficient for the next image data. The difference coefficient memory 180 is preferably the block diagram shown in FIG. This is similar to that described in the third embodiment.

(Sixth Embodiment) A sixth embodiment of the present invention will be described with reference to FIG. FIG. 7 shows an image recording apparatus for explaining the sixth embodiment of the present invention. External input terminal 330, changeover switches 350, 360, image recompression device 300
And a recorder 340. Image recompression device 30
As 0, any of FIG. 3, FIG. 5, and FIG. 6 may be used.

First, the bit stream is input terminal 330.
Input from. When recording in the normal mode, the selector switch 350 is set to the terminal 310 and the selector switch 36
0 is connected to terminal 320. In this case, the bit stream is not processed and is sent to the recorder 340 as it is and recorded on the recording medium (magnetic tape, optical disk, etc.). On the other hand, in the long-time recording mode, the changeover switch 35
0 to terminal 100 and changeover switch 360 to terminal 200
Connect to. When connected in this manner, the device has substantially the same configuration as the device described in the third to fifth embodiments, and the bitstream recompressed by the image recompression device 300 is recorded by the recorder 340. By switching the switch in this way, it becomes possible to perform standard recording or long-term recording of the bitstream.

[0038]

As is apparent from the above description, according to the image recompression method and the image recompression apparatus of the present invention, it is not necessary to completely solve the bitstream to reproduce the image, and the input bitstream is not It is possible to recompress to the bit rate. Therefore, there is almost no delay. Furthermore, since the prediction error of the current image is corrected by the error signal generated by recompressing the past image and then the image is recompressed, the propagation of the error due to the recompression can be prevented,
It is possible to reduce deterioration of image quality.

[Brief description of drawings]

FIG. 1 is a flowchart for explaining a first embodiment of an image recompression method according to the present invention.

FIG. 2 is a flowchart for explaining a second embodiment of the image recompression method according to the present invention.

FIG. 3 is a block diagram of an image recompression device for explaining a third embodiment of the present invention.

FIG. 4 is a block diagram of a difference coefficient memory in FIG.

FIG. 5 is a block diagram of an image recompression device for explaining a fourth embodiment of the present invention.

FIG. 6 is a block diagram of an image recompression device for explaining a fifth embodiment of the present invention.

FIG. 7 is an explanatory diagram of an image recording apparatus for explaining a sixth embodiment of the present invention.

FIG. 8 is a flowchart for explaining a conventional image recompression method.

FIG. 9 is a block diagram of a conventional image recompression device.

[Explanation of symbols]

 110 Variable Length Decoder 130 First Inverse Quantizer 140 Quantizer 150 Variable Length Encoder 160 Multiplexer 170 Second Inverse Quantizer 180 Difference Coefficient Memory

─────────────────────────────────────────────────── ─── Continuation of the front page (51) Int.Cl. ⁶ Identification number Office reference number FI technical display location H04N 7/32 H04N 7/137

Claims (10)

[Claims] 1. The image data quantized and compressed with a first quantization parameter is inversely quantized with the first quantization parameter to generate first inverse quantized image data, and the first inverse quantized image data is generated. From the quantized image data, the correction signal generated by the past image data is subtracted, quantized and output with the second quantization parameter, and dequantized with the second quantization parameter, and the second dequantization. Image data is generated, a difference signal is generated from the second dequantized image data and the first dequantized image data, and the difference signal is used as a correction signal of the next image data. Image recompression method. 2. A method of adding a correction signal generated by the past image data to the second dequantized image data and then generating a difference signal from the first dequantized image data. The image recompression method according to claim 1, wherein the image recompression method is used. 3. A second transform coefficient block is generated by subtracting a correction coefficient generated by past image data from a first transform coefficient block included in image data that has been orthogonally transformed and compressed. A predetermined number of coefficients are extracted from the second transform coefficient block to form and output a third transform coefficient block, and from the third transform coefficient block and the first transform coefficient block. An image recompression method, wherein a difference coefficient block is generated and the difference coefficient block is used as a correction coefficient for the next image data. 4. A difference coefficient block is generated from the first transform coefficient block after adding a correction coefficient generated by the past image data to the third transform coefficient block. The image recompression method according to claim 3. 5. A variable length decoder, a first dequantizer, and
A first adder, a quantizer, a variable length encoder, a multiplexer, a second dequantizer, second and third adders, and a difference coefficient memory, The converted image data is input, the variable length decoder decodes the first quantization parameter and the first quantized coefficient, and the decoded image data is sent to the first dequantizer. Data other than the quantization parameter and the first quantized coefficient is sent to the multiplexer, and the first dequantizer inverses the first quantized coefficient with the first quantization parameter. Quantize to generate a first dequantized coefficient, in the first adder, from the first dequantized coefficient,
Correction coefficients generated by past image data stored in the difference coefficient memory are added, input to the quantizer, quantized with a second quantization parameter, and a second quantized coefficient is obtained. The second quantization parameter and the second quantized coefficient are input to the variable length encoder, converted into a variable length code, sent to the multiplexer, and multiplexed with the other data. While outputting as, in the second dequantizer, by dequantizing the second quantized coefficient with the second quantization parameter, to generate a second dequantized coefficient, the A second adder adds the correction coefficient generated by the past image data to the second dequantized coefficient, and then the third adder adds the first dequantized coefficient. And a difference coefficient is generated from Storage, and the image recompression apparatus, characterized by using the difference coefficient as the correction coefficient of the next image data. 6. The difference coefficient memory comprises an inverse orthogonal transformer, a memory, and an orthogonal transformer, wherein the difference coefficient is input to the inverse orthogonal transformer, inverse orthogonal transformed, and stored in the memory. The difference coefficient is taken out from the memory based on the motion information input from the variable length decoder, and the difference coefficient is orthogonally transformed by the orthogonal transformer and sent to the first adder. The image recompression device according to claim 5. 7. A variable length decoder, a first inverse quantizer,
First inverse orthogonal transformer, first adder, orthogonal transformer, quantizer, variable length encoder, multiplexer, second inverse quantizer, second inverse orthogonal transformer , A second and a third adder, and a differential signal memory, and inputs encoded image data, and the variable length decoder inputs the first quantization parameter and the first quantized Decoding the quantized coefficient and sending it to the first dequantizer, sending the other data other than the first quantization parameter and the first quantized coefficient to the multiplexer, the first dequantizer At, to generate a first dequantized coefficient by dequantizing the first quantized coefficient with the first quantization parameter, in the first inverse orthogonal transformer, the first The inverse quantization coefficient is converted into a first spatial signal, and the first adder adds The correction signal generated by the past image data stored in the differential signal memory is subtracted, orthogonally transformed by the orthogonal transformer and then input to the quantizer, and quantized by the second quantization parameter to Two quantized coefficients are generated, the second quantization parameter and the second quantized coefficient are input to the variable length encoder, converted to a variable length code and sent to the multiplexer, While multiplexed with other data and output, in the second dequantizer,
Dequantize the second quantized coefficient with the second quantization parameter to generate a second dequantized coefficient, and in the second inverse orthogonal transformer, the second dequantized A coefficient is converted into a second spatial signal, and in the second adder, after adding the correction signal generated by the past image data to the second spatial signal, the third adder An image recompression device, wherein a differential signal is generated from the first spatial signal, stored in the differential signal memory, and the differential signal is used as a correction signal for the next image data. 8. A variable length decoder, a first adder, a coefficient selector, a variable length encoder, a multiplexer, second and third adders, and a difference coefficient memory. Then, the encoded image data is input, the first transform coefficient block is decoded by the variable length decoder and sent to the first adder, and other data other than the first transform coefficient block is sent. Sent to the multiplexer, the first adder subtracts the correction coefficient generated by the past image data stored in the difference coefficient memory from the first transform coefficient block, Generating a transform coefficient block, extracting a predetermined number of coefficients from the second transform coefficient block by the coefficient selector to form a third transform coefficient block, the variable length coding The third transform coefficient block To a variable length code, sent to the multiplexer, multiplexed with the other data and output, and sent to the second adder, the second adder, the third transform coefficient block In addition, after adding the correction coefficient generated by the past image data, in the third adder, to generate a difference coefficient from the first conversion coefficient block, and stored in the difference coefficient memory, An image recompression device, wherein the difference coefficient is used as a correction coefficient for the next image data. 9. The difference coefficient memory comprises an inverse orthogonal transformer, a memory, and an orthogonal transformer, wherein the difference coefficient is input to the inverse orthogonal transformer, inverse orthogonal transformed, and stored in the memory. Based on the motion information input from the variable length decoder, a difference coefficient is extracted from the memory, the difference coefficient is orthogonally converted by the orthogonal transformer, and the difference coefficient is transmitted to the first adder. The image recompression device according to claim 8. 10. An external input terminal and claims 5 to 9.
The image recompression device according to any one of 1 to 3, and a recorder, wherein encoded image data is input from the external input terminal, and in the normal recording mode, the encoded image data is directly recorded. An image recording apparatus, characterized by inputting to a recording device and recording, and in a long-time recording mode, inputting the coded image data to the image recompressing device, recompressing the image data, and then recording to the recording device. .