JP2002218471A - Method for converting image encoded data rate and device for converting image encoding rate - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an image encoded data rate converting method which is easily realized without decoding an encoded image signal up to the level of image data, in which delay time is small and image quality deterioration is low. SOLUTION: In this image encoded data rate converting method, a signal separating means 13 divides a signal encoded by motion compensation prediction and conversion encoding into an encoded image signal and encoded information, an image rate converting means 20 generates updated image data by leaving one or more quantization coefficients among a plurality of quantization coefficients of each pixel block of the encoded image signal and eliminating the other quantization coefficients, and a signal coupling means 15 obtains a stream updated by being coupled with the encoded information obtained by the signal separating means.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to rate conversion of coded image data encoded at a predetermined transfer rate.
In particular, the present invention relates to a rate conversion method for image encoded data and an image encoding rate conversion apparatus capable of performing rate conversion with a simple configuration in a device that records, transmits, or displays encoded moving images.

[0002]

2. Description of the Related Art As a technology for encoding moving picture signals such as television signals with high efficiency, an MPEG-standard is used as an international standard.
2 (moving picture experts group -2) is ISO / I
EC (International Organization for Standardizati
on / International Electrotechnical Commission).

[0003] MPEG-2 divides a "frame" image constituting a moving image into 16 x 16 pixel blocks called "macroblocks".
A motion amount called a “motion vector” is obtained between a reference image and a coded image separated by a predetermined number of frames in the future or the past in time, and the coded image is encoded from the reference image based on the motion amount. A DCT (Discrete Cosine) technique, which is one of orthogonal transform techniques, is applied to an "motion compensated prediction"
Transform: Discrete cosine transform), which converts image information into a frequency information amount, and then performs compression encoding by obtaining only visually significant information. It is stipulated on the basis of elemental technologies for chemical conversion.

[0004] The transfer rate of a signal image-encoded by the technique specified in this manner is set within a fixed transfer rate at which the transmission capacity per unit time is almost constant and a fixed transfer rate. Two types of setting methods are used: a variable transfer rate at which the transfer is performed with a variable transmission capacity.

[0005] Signals encoded using the MPEG-2 standard are used in digital broadcasts, recording media such as DVDs, and communication paths such as ATM (Asynchronous Transfer Mode). The bit rate indicating the transmission capacity of the bit stream per unit time can be freely selected within a certain range according to the application.

As described above, since coded signals having different transfer rates are used depending on applications, the coded signals having different transfer rates are transmitted at a rate of, for example, 7 Mbps (M bit / bit).
Second) digital broadcast signal cannot be recorded on a digital recorder set to a recording rate of, for example, 4 Mbps.

At this time, there is a method of recording by setting the recording capacity of the recorder to a large bit rate in accordance with the signal rate of digital broadcasting. In this case, it is necessary to secure a large recording area on the recording medium of the recorder. Therefore, a problem such as a reduction in recording time occurs, which is not preferable.

Therefore, a 7 Mbps bit rate signal is
The signal is converted into a signal of a new bit rate of Mbps and then recorded on the recorder, so that a predetermined recording time is secured.

The rate conversion is performed when a bit stream broadcast at a fixed transfer rate is recorded at a variable transfer rate, or a bit stream broadcast at a variable transfer rate opposite thereto is recorded at a fixed transfer rate. There is also a need for rate conversion processing when switching between such fixed and variable bit rate setting methods.

As described above, as a method of performing a rate conversion and generating a bit stream of another transfer rate from a bit stream of a certain transfer rate, conventionally, a supplied bit stream is decoded to a pixel level by a decoder and decoded. After generating an image, the generated decoded image is supplied to an encoder to perform encoding again, so-called “re-encoding”.
Method has been used.

However, re-encoding requires both a decoder and an encoder. In addition, an image memory for temporarily storing decoded images is required for encoding, and the circuit scale is increased. There is a problem that the encoding and re-encoding processes are delayed.

As a rate conversion method for solving the problem, variable length decoding (VLD; decoding of variable l) is performed without decoding an input bit stream to a pixel level.
DCT obtained by performing the encoding and the inverse quantization
A rate conversion method for re-quantizing coefficients with a quantization scale of a different value to obtain a bit stream of a required bit rate is disclosed in Japanese Unexamined Patent Application Publication No. 7-312756, entitled "Information Conversion Circuit for Compressed Video Code Signal , Apparatus and methods ".

FIG. 16 shows a configuration of a rate conversion device in a conventional example having such a configuration. In the figure, a supplied 4 Mbps bit stream is separated into an image signal and other signal parts by a data separation circuit 81, and the image signal is converted into a V-length signal by an inverse VLC (variable length coding) circuit 82.
The LC-decoded signal is decoded, and the signal obtained by the decoding is inversely quantized by an inverse quantizer 83, and then quantized by a quantizer 85 controlled by a bit rate control circuit 84. The signal obtained by the conversion is subjected to VLC by a VLC circuit 86, the signal obtained by the VLC is combined with a signal separated by a data separation circuit 81 by a combining circuit 87, and the combined signal is temporarily stored in a buffer circuit 88. The stored and temporarily stored signal whose transfer rate has been converted to 2 Mbps is read out from the buffer circuit 88 as necessary, and supplied as an output signal from the rate converter.

In this manner, a signal whose rate has been converted by inverse quantization and requantization can be obtained.
As another rate conversion method, when the bit rate after the rate conversion is lower than that before the conversion, the variable length code corresponding to the DCT coefficient portion after the quantization of the input bit stream is adjusted to reduce the code length. A rate conversion method for shortening and obtaining a bit stream of a required bit rate is disclosed in Japanese Patent Application Laid-Open No. H11-317942, entitled "Image Encoding Apparatus".

[0015]

By the way, MPEG-
In a signal obtained by encoding according to the two schemes or the like, a bit stream signal is formed by prefixing information on the encoding parameter to the encoded signal as a header signal.

Therefore, when the coding parameter is changed by the rate conversion processing of the coded signal, it is necessary to update the contents of the header signal in accordance with the change of the coded signal.

The update of the header signal is performed by the method of directly adjusting the variable length code of the DCT coefficient portion in the above-described conventional example and performing the rate conversion, for example, when adjusting the DCT coefficient of a non-intra macro block. For example, when the number of DCT coefficients is changed by rate conversion and the number is made zero, the CHP (CodedBl
It is necessary to update a header signal such as an ock pattern.

However, since the header signal is updated only after all the image signals related to the header signal have undergone rate conversion processing, the header signal is not determined. It is necessary to store the information, which causes a problem that the configuration of the apparatus is complicated and that a delay time is generated due to a rate conversion process.

In order to prevent such a problem from occurring, a method is conceivable in which the rate conversion is performed to the extent that the header signal is not changed, but it is particularly difficult to convert the signal to a low rate. These rate conversion methods are effective, for example, if the processing of quantizing the coefficient of the block in which the DCT coefficient should be left to obtain a stream of coarse coefficients or reducing the coefficient, the image quality deteriorates greatly. Had not been used for

Therefore, according to the present invention, the transfer rate is converted while maintaining the same data structure for the coding method of the data structure in which image data such as MPEG-2 is divided into macroblocks and coding is performed. That is, each divided macro block has the same macro block type and CBP (coded block pattern) as before conversion.
To convert the signal to

For this reason, for a block in which DCT coefficients coded before conversion exist, at least one coefficient remains after conversion, and when rate conversion is performed by requantization, the quantization before conversion is performed. Update of quantization scale The quantization scale is updated only at the same position as the macroblock position.

In the rate conversion performed in this manner, the data to be converted includes only the DCT coefficient part when operating the variable length code of the DCT coefficient part, and the DCT coefficient part when performing the rate conversion by requantization. To provide a rate conversion processing method capable of simplifying the processing associated with the rate conversion, such as being able to use only the quantization scale code and the quantization scale code, with less image quality degradation, and a configuration of a rate conversion processing apparatus equipped with the processing method. It is assumed that.

[0023]

The present invention comprises the following means 1) to 7) to solve the above-mentioned problems. That is,

1) A moving image signal is divided into a plurality of image data for each pixel block of a predetermined size, a motion vector amount is obtained for each of the divided pixel blocks, and a motion vector is calculated based on the obtained motion vector amount. Perform compensation prediction, perform orthogonal transformation of the image data of the pixel block, quantize the data obtained by the orthogonal transformation to obtain quantized data,
A coded image signal is generated by performing variable-length coding on the obtained quantized data, and information on coding parameters related to generation of the coded image signal is obtained as coding information.
A header signal including the obtained encoded information is generated, and the generated encoded image signal and header signal are combined to obtain first encoded data compressed and encoded at a first transfer rate. And an image obtained by supplying the obtained first encoded data and converting the supplied first encoded data into second encoded data compressed and encoded at a second transfer rate. A rate conversion method for encoded data, wherein a first step (1) of obtaining the encoded image signal and the encoded information from the first encoded data
3) and image data for each pixel block obtained by orthogonally transforming the coded image signal based on the coded information obtained in the first step, and relates to the obtained image data. A second step (20) of generating updated image data by deleting one or more of the plurality of quantized coefficients of the quantized data and deleting other quantized coefficients; and an update generated by the second step. A third step (15) of combining the image data with updated encoded information based on the encoded information obtained in the first step to obtain the second encoded data. Characteristic encoded image data rate conversion method.

2) The updated image data in the second step may include one or more of a plurality of quantized coefficients obtained by re-quantizing the image data with the updated quantization scale, and other quantized coefficients. The rate conversion method for image-encoded data according to item 1), wherein the rate conversion is performed by deleting the conversion coefficient.

3) The updated image data in the second step is sequentially deleted from a plurality of quantized coefficients of the quantized data relating to the image data, from a quantized coefficient relating to a higher-order orthogonal transform coefficient. The rate conversion method for encoded image data according to the item 1), wherein the rate conversion is performed.

4) The updated image data in the second step is a quantized coefficient which gives a maximum absolute value related to an orthogonal transform coefficient among a plurality of quantized coefficients of the quantized data related to the image data. , The rate conversion method of the encoded image data according to the item 1).

5) The updated image data in the second step is a quantum that gives the shortest code length by performing the variable-length encoding among a plurality of quantized coefficients of the quantized data relating to the image data. The rate conversion method for image-encoded data according to item 1), wherein the rate conversion is performed so as to leave the conversion coefficient.

6) The moving image signal is divided into a plurality of image data for each pixel block of a predetermined size, a motion vector amount is obtained for each of the divided pixel blocks, and a motion vector is calculated based on the obtained motion vector amount. Perform compensation prediction, perform orthogonal transformation of the image data of the pixel block, quantize the data obtained by the orthogonal transformation to obtain quantized data,
A coded image signal is generated by performing variable-length coding on the obtained quantized data, and information on coding parameters related to generation of the coded image signal is obtained as coding information.
A header signal including the obtained encoded information is generated, and the generated encoded image signal and header signal are combined to obtain first encoded data compressed and encoded at a first transfer rate. And the obtained first encoded data is supplied, and the image encoded data obtained by converting the first encoded data into the second encoded data compressed and encoded at the second transfer rate is obtained. The rate conversion method, wherein the encoded image signal from the first encoded data,
And a first step (13) of obtaining the encoded information;
Based on the encoded information obtained in the first step,
Image data for each pixel block obtained by orthogonally transforming the coded image signal is obtained. Instead of a plurality of quantized coefficients of quantized data related to the obtained image data, a two-dimensional V
A second step (20) of generating updated image data by using a coefficient value having a short LC code length and having little effect on the image quality, the updated image data generated by the second step, A third step (1) of obtaining updated second encoded data by combining updated encoded information based on the encoded information obtained in step 1;
5) A method for converting the rate of encoded image data, which comprises:

7) A moving image signal is divided into a plurality of image data for each pixel block of a predetermined size, a motion vector amount is obtained for each of the divided pixel blocks, and a motion vector is calculated based on the obtained motion vector amount. Perform compensation prediction, perform orthogonal transformation of the image data of the pixel block, quantize the data obtained by the orthogonal transformation to obtain quantized data,
A coded image signal is generated by performing variable-length coding on the obtained quantized data, and information on coding parameters related to generation of the coded image signal is obtained as coding information.
A header signal including the obtained encoded information is generated, and the generated encoded image signal and header signal are combined to obtain first encoded data compressed and encoded at a first transfer rate. At the same time, the obtained first coded data is supplied, and the image coding rate obtained by converting the first coded data into the second coded data compressed and coded at the second transfer rate is obtained. A signal separating means (13) for obtaining said coded image signal and said coded information from said first coded data, and for each pixel block obtained by orthogonally transforming said coded image signal. Image data conversion means for obtaining updated image data and deleting at least one of a plurality of quantized coefficients of the quantized data relating to the obtained image data and deleting other quantized coefficients to generate updated image data (20), Signal combining means for combining the updated image data generated by the image data converting means with updated coding information based on the coding information obtained by the signal separating means to obtain the second coded data An image coding rate conversion device characterized by comprising (15).

[0031]

BEST MODE FOR CARRYING OUT THE INVENTION Hereinafter, a preferred embodiment of an image encoding data rate conversion method and an image encoding rate conversion apparatus according to the present invention will be described. FIG.
This is a configuration of an image encoding rate conversion apparatus equipped with an image encoded data rate conversion method for performing transfer rate conversion while keeping the data structure of the encoded signal the same, and will be described below with reference to the drawings.

An image coding rate converter 10 shown in FIG.
Is a device that performs rate conversion by inverse quantization and requantization, and the device is a DEMUX (de-multiplexin).
g) unit 11, input code amount counter 12, header separation unit 1
3, header data storage unit 14, header combining unit 15, output code amount counter 16, re-MUX (multiplexing) unit 1
7, and an image rate conversion unit 20 that performs rate conversion by inverse quantization and requantization.

Next, the operation of the image coding rate converter 10 configured as described above will be described. First, at a predetermined data rate, MPEG-2 (moving picture experts)
The bit stream encoded according to the group-2) method includes an image signal encoded by the DEMUX unit 11, an audio signal attached to the image signal, auxiliary signals such as character information, and a method of reproducing these signals. And other signals related to the MPEG-2 system for performing control, synchronization and the like.

The other signal portions obtained separately are supplied to the re-MUX unit 17, and the image signals obtained separately are supplied to the input code amount counter 12, where the code amount of the image signal is counted. , A part of the signal obtained by the counting is supplied to the image rate conversion unit 20, and the other part is supplied to the header separation unit 13.

The header separation section 13 separates the data of the layers above the slice layer such as the sequence header and the picture header from the supplied image signal, and outputs the separated signal to the header data storage section 14. And temporarily stored therein. The signal temporarily stored therein is supplied to the header combining unit 15 as necessary, and the bit rate information before conversion recorded in some other sequence headers and the encoding. The converted image signal is converted to an image rate
Supplied to

In the image rate conversion section 20, the header separation section 13 uses the pre-conversion bit rate information supplied from the input code amount counter 12 and the post-conversion bit rate information supplied from the output code amount counter 16. Perform image rate conversion of the supplied encoded image signal,
The converted image signal obtained by the conversion is supplied to the header combining unit 15 and combined with the header signal supplied from the header data storage unit 14. The combined signal is supplied to the output code amount counter 16. You.

For the signal supplied to the output code amount counter 16, the bit rate of the encoded signal is measured to obtain the above-mentioned converted bit rate information. The audio signal which is supplied to the MUX unit 17 and added to the above-mentioned image signal and other signal parts related to the MPEG-2 system are added and supplied from the image encoding rate conversion device 10 as an MPEG-2 stream output signal. Is done.

The MPEG-2 stream whose coding rate has been converted is supplied as an output signal from the image coding rate converter 10 in this manner. Next, an image rate converter, which is a main part of the converter, is provided. About the operation of 20,
This will be described together with the encoding operation performed by the MPEG-2 encoding method.

FIG. 2 shows the configuration of an MPEG-2 encoder that encodes a moving picture signal specified by MPEG-2.

The MPEG-2 encoder 60 includes a subtractor 61, a DCT (Discrete Cosine Transform) unit 62,
Quantizer 63, code amount controller 64, variable length encoder 6
5, buffer 66, inverse quantizer 71, IDCT (Invers
e Discrete Cosine transform) unit 72, adder 73,
It comprises a frame memory 74 and a motion compensation predictor 75.

The operation of the MPEG-2 encoder 60 configured as described above will be described. First, the supplied moving image signal is supplied to one input terminal of a subtractor 61 and also to a motion compensation predictor 75, where a moving vector of the moving image signal is obtained, and based on the obtained moving vector, The motion prediction signal is supplied to one input terminal of an adder 73, and the motion prediction signal is supplied to a subtractor 61 as a subtraction input signal.

From the subtracter 61, a difference signal obtained by subtracting the motion prediction signal from the supplied moving image signal is converted into a DCT signal.
The DCT unit 62 performs a discrete cosine transform of the supplied difference signal, and a cosine frequency component obtained by the cosine transform is supplied to a quantizer 63.

In the quantizer 63, a code amount control unit 64
Of the cosine frequency components supplied from the DCT unit 62 based on the control signal supplied from the DCT unit 62, a frequency component effective for image encoding is obtained as a quantized signal, and the obtained quantized signal has a variable length. The signal is supplied to the encoder 65 and also supplied to the inverse quantizer 71.

The quantized signal supplied to the inverse quantizer 71 is inversely quantized to obtain an approximate cosine frequency component. The obtained approximate cosine frequency component is supplied to the IDCT unit 72. Thus, an approximate image difference signal is obtained.

The obtained approximate image difference signal is supplied to an adder 73, and an approximate image signal is decoded by being added to the above-described motion prediction signal, and the decoded image signal is decoded. Are stored in the frame memory 74,
The image signal stored in the frame memory 74 is used to generate a motion-compensated predicted image for the frame image of the next incoming moving image signal.

The video signal is coded while the motion compensated prediction image is being generated in this manner. The quantized signal supplied to the above-described variable length coder 65 is subjected to variable length coding, The signal obtained by the long encoding is temporarily stored in a buffer 66 and supplied as an encoded bit stream, and a part of the signal temporarily stored in the buffer 66 is supplied to a code amount control unit 64. The code amount control unit 64 detects the code amount of the generated bit stream, and generates a control signal for controlling the quantization roughness of the quantizer 63 based on the detected code amount information. The quantizer 63 generates an encoded signal at a predetermined bit rate.

As described above, the supplied moving image signal is subjected to image coding using the motion compensation prediction method and the transform coding method. Next, the image signal to be subjected to the image encoding is formed into a coded picture configuration for each predetermined frame image, and motion compensation prediction is performed.
A configuration of a predicted frame image for which motion compensation prediction is performed in the scheme will be described.

FIG. 3 shows a coded picture structure for explaining the relation between picture images for which motion vectors are obtained and motion compensation is performed for the motion compensation prediction performed by the coding device 60.

The structure of an encoded picture based on the motion compensated prediction shown in FIG. 1 includes an I (Intra-coded) picture (intra-frame encoding) and a P (Predictive).
-coded) picture (forward predictive coding), and B
(Bidirectionally predictive-code
d) It is composed of a combination of three types of pictures having different motion prediction methods, called pictures (bidirectional predictive coding).

The frame-based image signal composed of such a motion prediction structure is composed of a normal video signal as an image signal of 30 frames per second and a movie material as an image signal of 24 frames per second. Transform coding of a signal will be described.

FIG. 4 shows an image division method related to the transform coding. In the figure, the frame image has a width of 16
This indicates that image compression processing is performed by dividing the image into a horizontally long image called a slice composed of book scanning lines.

As a pixel unit for the image compression processing, a slice is divided into 16 pixels in the horizontal direction, and a pixel unit of 16 pixels × 16 scanning lines (pixels) obtained by dividing the slice is called a macroblock. It is a pixel unit.

The macro block which is a pixel unit includes four luminance signal blocks of 8 × 8 pixels,
Transform coding is performed by dividing into a total of six blocks including a color difference signal block in which each of blue and red is 8 pixels × 8 pixels.

The image data thus transformed and encoded includes a slice layer composed of encoded data relating to image data composed of slices, a macroblock layer relating to data for each macroblock, and a macroblock. Are coded data composed of data of each layer such as a block layer composed of DCT coefficient data obtained by quantizing the six blocks constituting.

The data of each layer configured as described above is as follows:
Data that is transmitted following a header such as a sequence header or a picture header is called a bit stream.

FIG. 5 shows the MPEG transmitted in this manner.
2 shows a data structure of a bit stream in -2. In the figure, the data to be transmitted is shown to be composed of header data, slice layer data, macro block layer data, and block layer data.

FIG. 6 shows a data structure in the macro block layer. The data in the macroblock layer is
Macro block escape, macro block address increment, macro block type, motion type, DCT type, Q scale code, motion vector, and CBP (coded block pattern) are arranged in this order.

The macroblock types in these arrays indicate the prediction mode, the presence / absence of a quantized scale code, and the presence / absence of a coded coefficient for each picture type represented by I, P, and B. The updated scale code is transmitted such that the updated value is transmitted only when the quantization scale is updated in the macroblock in the slice.

Next, a macroblock type VLC (variable length) for each picture given by the I, P, B
h coding) table. FIG. 7 shows the macroblock type VL for I-pictures in those pictures.
This is a C table. When the code is “1”, the description indicates that the code is an intra code, and when the code is “01”, it indicates that the quantization scale code is arranged later. Is shown.

FIG. 8 shows a macroblock type VLC table for a P picture, which shows a state in which motion prediction coding (MC: Motion Compensation) is performed by a description for the code, and a quantization. This shows the status of the scale code (Quant).

FIG. 9 is a VLC table of a macroblock type for a B picture.
ward) shows the state of the prediction coded data regarding prediction from an image and prediction from a backward (Bwd: backward) image.

In this way, the macroblock type information is arranged and transmitted as a bit stream.
The CBP is arranged at the end of the macro block layer, and indicates whether or not the quantized DCT coefficient data of the subsequent block layer exists in each of the six blocks.

When the macro block type is an intra macro block which is intra-coded image data, at least one coefficient data is present in every block. The last CBP is made nonexistent.

When coefficients do not exist in all blocks in one macro block, the macro block is set to "Not Coded".

Further, in the case of frame structure coding, when the motion type (motion prediction type) is frame prediction, and all the motion vectors are "0" in P pictures (this is called "No MC"), When "Not Coded" is reached, the macroblock is skipped, that is, not coded,
It can be.

Similarly, the motion type (motion prediction type) is frame prediction, the prediction direction of the B picture such as forward prediction, backward prediction or bidirectional prediction, and the macroblock immediately before the slice having the same motion vector. If they are the same, the macroblock can be skipped, that is, not encoded, as “Not Coded”.

The data of the macro block layer is arranged in this way, and the data of the block layer is arranged after the macro block layer. Next, the data structure of the block layer will be described. FIG. 10 shows a data structure in the block layer.

In the figure, a block layer has a DC (direct current) coefficient of an image obtained by performing a DCT operation on image data of an intra macroblock and AC which is a cosine frequency component.
(AC) coefficients and the E transmitted after the completion of those coefficients
They are arranged in the order of OB (end of block).

The DC after quantization in the block layer
The coding of the T coefficient data is different between the intra macroblock and the other macroblocks. In the case of the intra macroblock, there is only the DC coefficient which is the first coefficient, which is different from the AC coefficient in the VLC (variab).
le length coding (variable length coding).

That is, the coefficients other than the DC coefficients of the intra macroblock and the coefficients of the other macroblocks are rearranged in the specified order, and the number (run) of leading zero coefficients in that order and the non- Zero coefficient value (level)
Are checked and encoded by means of a two-dimensional VLC table showing the runs and levels in one set.

In the coding by the two-dimensional VLC table, a code called EOB (end of block) is added when the non-zero coefficient is lost, and the coding of the block is completed.

The bit stream generated by arranging the moving picture signal by MPEG-2 and arranging it by a predetermined method for transmitting the coded image data has been described.

Next, a description will be given of a method of converting the coded signal generated at a predetermined bit rate into a signal having another desired bit rate.

The conversion is performed, for example, when recording a received signal digitally broadcast by the MPEG-2 system into a digital VTR by converting the signal to a bit rate smaller than the broadcast so as to increase the recording time. It is.

When the bit rate is converted from a digital signal to an analog signal and then to a digital signal again, the processing time for the conversion becomes long. For example, recording and reproduction of the supplied signal The delay time becomes an obstacle when simultaneous monitoring of recording signals is performed simultaneously, and when a digital signal is converted into an analog signal and then converted again into a digital signal, the image quality deteriorates due to the conversion. An object of the present invention is to realize a configuration of an image coding rate conversion device that does not deteriorate image quality due to reconversion, and will be described below with a plurality of embodiments.

FIG. 11 shows the configuration of an image coding rate converter according to the first embodiment. The image coding rate conversion device 10 is a device that performs rate conversion by inverse quantization and requantization.

Then, the image coding rate converter 1
0 indicates a DEMUX (de-multiplexing) unit 11, an input code amount counter 12, a header separation unit 13, a header data storage unit 14, a header combining unit 15, an output code amount counter 16,
Re-MUX (multiplexing) unit 17, variable-length decoding unit 21,
It comprises an inverse quantization / requantization unit 22, a code amount control unit 23, a coefficient selection unit 24, and a variable length coding unit 25.

Next, the operation of the image coding rate converter 10 configured as described above will be described.

First, at a predetermined data rate, MPEG-2
The bit stream encoded according to the method is a DEMU
The image signal encoded by the X section 11 and the audio signal attached to the image signal and other signal parts related to the MPEG-2 system are separated, and the other signal parts obtained by the separation are While being supplied to the re-MUX unit 17, the image signal is also supplied to the input code amount counter 12, where the code amount of the image signal is counted, and a part of the counted signal is supplied to the code amount control unit 23. The other part is supplied to the header separating unit 13.

In the header separating section 13, data of a layer higher than the slice layer such as a sequence header and a picture header is separated from the supplied image signal, and the separated signal is supplied to a header data storage section 14. The bit rate information before conversion, which is temporarily stored and recorded in the sequence header which is another part, is stored in the code amount control unit 23.
And another part is supplied to the variable length decoding unit 21.

The signal supplied to the variable-length decoding unit 21 from which the header data is separated is subjected to DCT by the variable-length decoding unit 21.
(Discrete cosine transform) The DCT coefficient part is divided into a coefficient part and other parts, and is subjected to variable length decoding.
The pre-transform quantization scale code obtained by decoding is supplied to the inverse quantization / requantization unit 22.

The slice layer obtained by decoding, the pre-conversion quantization scale code recorded in the macroblock layer, and the position information of the macroblock in the screen are supplied to the code amount control unit 23.

The code amount control unit 23 receives the pre-conversion bit rate information, the post-conversion target bit rate information, the pre-conversion quantized scale code, the macroblock position information, the input code amount counter 12 and the output code amount counter 16. Code amount information of each image before and after the supplied conversion, and inverse quantization
Code amount information before and after rate conversion of each macroblock counted by the requantization unit 22 is supplied.

The code amount control unit 2 to which these information are supplied
In 3, an average value of the code amount for each image with respect to the quantization scale before and after the conversion of each image is calculated based on the information, and a predetermined method for the obtained code amount, for example, By controlling the code amount by the method specified in Steps 1 and 2 of MPEG-2 Test Model 5, the quantization scale after conversion is determined, and the determined quantization scale after conversion is inverse quantization / requantization. Is supplied to the conversion unit 22.

The code amount control in the inverse quantization / requantization unit 22 can use an arbitrary quantization scale for each macroblock during normal encoding, that is, an arbitrary quantization scale code during rate conversion. However, here, the quantization scale after conversion can be changed only at the macroblock position where the quantization scale code exists in the data before conversion.

FIG. 12 shows how a quantized scale code exists before and after rate conversion. In the figure, the left-side diagram shows the state of the scale code before the rate conversion, and the right side shows the scale code after the rate conversion.

The left side of each figure shows the value of the slice layer and the right side shows the value of the macro block layer, and the halftone dots indicate the positions where the quantization scale codes exist.

As described above, the quantized scale code after the rate conversion exists at a part of the position existing before the rate conversion, and the quantized scale after the conversion is inverse quantum scale. It is supplied to the quantization / requantization unit 22.

The inverse quantization / requantization unit 22 is supplied with the pre-transform quantization scale code and the DCT coefficient portion that have been variable-length decoded by the variable-length decoding unit 21, and the DCT coefficient portion is subjected to the pre-transform quantization. The data is inverse-quantized by the pre-conversion quantization scale obtained from the scale code, and the data obtained by the inverse quantization is requantized based on the converted quantization scale supplied from the code amount control unit 23. You.

That is, the inverse quantization and requantization processing in the inverse quantization / requantization unit 23 compares the supplied pre-conversion quantization scale and the supplied post-conversion quantization scale. The processing is performed according to the magnitude relation obtained. The inverse quantization and requantization processing includes the following three cases.

The first case is when the pre-transform quantization scale is equal to the post-transform quantization scale. In this case, the inverse quantization and requantization are not performed, and the DCT coefficient portion before the transformation is directly transformed. It is supplied as a later DCT coefficient part.

The second case is a case where the pre-conversion quantization scale is larger than the post-conversion quantization scale. In this case, for each DCT coefficient before conversion, the pre-conversion quantization scale is changed to the post-conversion quantization scale. And the updated DCT coefficient value obtained by the multiplication is supplied as the transformed DCT coefficient.

In the second case, since the DCT coefficient value updated by the rate conversion processing becomes large, the data structure is not kept the same because the coefficient value becomes 0 and the DCT coefficient value is deleted. Therefore, there is no need to perform processing such as coefficient selection for the updated DCT coefficient value.

In the third case, the pre-transform quantization scale is smaller than the post-transform quantization scale. In this case, the pre-transform quantization scale is converted to the post-transform quantization scale for each DCT coefficient before the transform. The updated DCT coefficient value is obtained by multiplying by the value of the ratio divided by the quantization scale. In this case, the updated DCT coefficient value is a small value.
In some cases, a block in which all updated DCT coefficients become 0 may occur.

Then, the updated DCT coefficient values are all 0.
In order to prevent that the data structure is not kept the same by being deleted in that block,
For the block whose updated coefficient value has become 0, the DCT coefficient before conversion is supplied to the coefficient selection unit 24, where a coefficient selection process is performed to prevent all the DCT coefficients after conversion from becoming 0.

In the coefficient selection process performed by the coefficient selection unit 24, first, of the DCT coefficients before conversion which are supplied from the inverse quantization / requantization unit 22 before the inverse quantization, the coefficients to be left are determined.

The first coefficient determination method is to set the DCT coefficient value after conversion at that position to "1" or "-1" with respect to the maximum absolute value of the DCT coefficient before conversion before inverse quantization. Is "0". For a block having a plurality of maximum values, the coefficient value at the position where the code length becomes the shortest when encoded by two-dimensional VLC, and the code length at that time is the same. In some cases, the coefficient is determined by selecting the coefficient value at the position where the zero run is the shortest.

The second coefficient determination method is based on whether the coefficient block is a luminance signal or a blue or red color difference signal regardless of the coefficient value of the DCT before conversion, and furthermore, based on the coefficient distribution of the DCT coefficient before conversion. The position of a coefficient that has the least effect on the image quality of the subsequent image and that reduces the code length of the two-dimensional VLC is determined in advance, and the coefficient at the determined position is determined to be the position of the DCT coefficient after conversion. Make a decision.

In this manner, the coefficients to be left or replaced by the coefficient selection section 24 are determined by the first or second method. The coefficient selection information is supplied to the coefficient operation section 27. The coefficient operation unit 27 performs inverse quantization and
The signals supplied from the requantization unit 22 and requantized are supplied to the variable-length encoding unit 25 as DCT coefficients of blocks in which all of them do not become 0.

In this way, the DCT coefficients of the blocks that are not all 0 are supplied to the variable length coding unit 25.
If the supplied macroblock signal is an intra-macro-coded intra-macroblock signal, the data structure is not changed even if all DCT coefficients become zero. Such processing of the coefficient selection unit 24 may not be performed.

As described above, the variable-length encoding unit 25 is also supplied with variable-length decoded data that has not been supplied to the inverse quantization / requantization unit 22.

In the variable length coding unit 25, the quantization scale code in the supplied data is determined by the code amount control unit 23 at the position of the macroblock where the quantization scale code before conversion exists. After the conversion, the converted code is converted into a code indicating the quantization scale, and the converted code is replaced with the converted code. Then, the replaced code is subjected to variable-length encoding.

Note that, depending on the state of the code amount control, the quantization scale code existing immediately before the same slice and the updated quantization scale code have the same value, and the quantization scale code is actually unnecessary. Macroblocks also exist, but their presence is not a problem with respect to the standards defined by MPEG-2.

In the opposite case, there is a macroblock in which the pre-conversion quantization scale code does not exist, but it is not necessary to newly add a post-conversion quantization scale code to the macroblock.

That is, when the quantized scale code does not exist, the transformed DCT coefficient supplied from the inverse quantization / requantization unit 22 is subjected to variable length coding by the variable length coding unit 25, and the variable length code The signal obtained by the conversion is combined with the quantized scale code included in the other variable-length code data separated by the variable-length decoding unit 21 to form a header combining unit 1
5 to be supplied.

In the header combining section 15, only the updated new bit rate value of the header data temporarily stored in the header data storage section 14 is replaced with the converted value and is changed by the two-dimensional VLC section 28. The data is subjected to long coding, combined with the converted data obtained by data combining in the data combining unit 29, and supplied to the re-MUX unit 17.

Then, the re-MUX unit 17 uses the DEMU
A bit stream signal that is combined with a signal portion other than the image coded signal supplied from the X section 11 and obtained is provided as an output signal of the image coding rate conversion device 10.

The configuration of the image coding rate conversion apparatus according to the first embodiment in which the rate conversion is performed by inverse quantization and requantization and the operation of the apparatus have been described above.
In the example shown here, the inverse quantization and requantization are performed simultaneously by the inverse quantization / requantization unit 22 for simplification of data processing. The quantization may be separately performed by the requantization unit.

FIG. 13 shows an image coding rate conversion apparatus 1 which is a modification of the first embodiment in which inverse quantization and requantization are performed separately.
0a is shown.

In the case of the modified embodiment, after the inverse quantization section performs inverse quantization on all blocks having DCT coefficients before conversion, the requantization section sets all converted DCT coefficients to 0. The requantization is performed so as not to be performed.

Next, a description will be given of an image coding rate conversion apparatus according to a second embodiment in which a rate conversion process is performed without using inverse quantization and requantization.

FIG. 14 shows the configuration of an image coding rate conversion apparatus according to the second embodiment. That is, the image coding rate conversion device does not perform inverse quantization and requantization,
Rate conversion is performed by operating the DCT coefficient portion.

The image coding rate conversion device 10b according to the second embodiment is different from the image coding rate conversion device 10 according to the first embodiment in the operation of the image rate conversion unit 20b.
That is, since only the operations of the DCT coefficient separation unit 26, the coefficient operation unit 27, the DCT coefficient combination unit 28, and the code amount control unit 23 are different, other operations are simplified and the operation of the image coding rate conversion device 10b is simplified. explain.

The MPEG-2 stream supplied to the image coding rate converter 10b is supplied to the header separation unit 13
Is supplied to the DCT coefficient separating unit 26, which is one of the encoded signals from which the header data portion has been separated. The DCT coefficient separating unit 26 separates the DCT coefficient portion from the image data portion and obtains it. , The resulting D
The data of the CT coefficient portion is supplied to the coefficient operation unit 27.

Then, the other header data portion separated by the header separation unit 13 is supplied to the code amount control unit 23, where the bit amount information before conversion, the target bit rate information after conversion, The code amount information of each image before and after conversion supplied from the input code amount counter 12 and the output code amount counter 16, the code amount information before and after rate conversion of each block counted by the coefficient operation unit 27, and If necessary, quantization scale code information is supplied, and the code amount is controlled by a predetermined method based on the supplied information.

In the code amount control method, after the target code amount for each image is determined, the target code amount is
Divide by the number of macroblocks included per frame, determine a target code amount per macroblock based on the code amount obtained by division, and assign a target code amount to the macroblock based on the determined code amount per macroblock. A target code amount is assigned to each of the six blocks included.

The target code amount for each block allocated in this way is supplied to the coefficient operation unit 27, and the coefficient operation unit 27 calculates the coefficient of each corresponding block based on the allocated code amount. The coefficient is controlled by the coefficient operation unit 27 so that the code amount of each block becomes the target code amount.

In the code amount control by another method, the bit rate ratio of the portion related to the DCT coefficient code amount of the converted target code amount with respect to the code amount before the rate conversion for each block is obtained. The determined bit rate ratio is multiplied by the pre-conversion code amount to determine a target code amount after rate conversion for each block, and the determined target code amount information after conversion is supplied to the coefficient operation unit 27, where the target code amount is calculated. The operation of the DCT coefficient portion is performed as obtained.

The DCT in the coefficient operation unit 27 at that time
In the operation of the coefficient portion, when the target code amount after rate conversion for each block is larger than the code amount before conversion, the coefficient operation process is not performed, but the converted bit rate signal equal to the target value is output. When it is desired to obtain, for example, dummy data is inserted into the image data position of each slice or picture as needed to obtain a signal of a target transfer rate.

When the target code amount after conversion is larger than the code amount before conversion, the transfer rate may be adjusted by inserting dummy data for each slice or picture. In that case, bit rate information and DCT The part other than the coefficient part does not correspond to the precondition that the part is not changed.

Next, adjustment of the transfer rate when the target code amount after conversion is smaller than the code amount before conversion will be described.
In this case, the data amount of the DCT coefficient portion before the rate conversion encoded by the two-dimensional VLC is reduced, and an operation for obtaining data of the target code amount after the conversion is performed.

For this operation, for example, the two-dimensional VLC code arranged before the EOB is deleted one word at a time by a combination of a run and a level, and the converted target code amount reaches a predetermined value. Sometimes there is a way to stop that deletion.

The deletion is stopped for the last two-dimensional VLC code arranged before the EOB even if the converted target code amount has not reached a predetermined value. To do.

In this case, a code indicating the type of the block indicating whether the coefficient block supplied from the coefficient selecting section 24 is a luminance signal or a blue or red color difference signal, and a pre-conversion D
2 indicating a coefficient determined in advance based on the coefficient distribution of the CT coefficient
Select a dimensional VLC code and add EOB to the selected code
Is determined, and the determined code is used as the code of the DCT coefficient portion after the transformation of this block.

If the macroblock at this time is an intra macroblock for performing intra-frame coding, the coefficients may be all zero, so that two-dimensional VLC other than EOB is used.
All codes may be deleted.

In this way, the DCT coefficient is deleted from the high-frequency component, and the data of the target code amount after conversion in which the data amount of the DCT coefficient portion is reduced is obtained. Next, the data amount is reduced by another method. The method of reduction is described.

In this method, a two-dimensional VLC code before conversion is decoded to obtain DCT coefficient sequence data, and a coefficient is cut by replacing a part of the obtained DCT coefficient sequence data with “0”. This is a method of performing two-dimensional VLC encoding on DCT coefficient sequence data from which coefficients have been cut.

According to this method, the magnitude of the degree of influence on the image quality after conversion is determined in advance as the code amount ratio before and after the rate conversion of the part where the coefficient sequence is replaced with “0”, and the coefficient is not set to “0”. When the target code amount is not satisfied, the replacement is performed with the coefficient to be left supplied from the coefficient selection unit 24.

In the coefficient selecting section 24, the coefficient operating section 27
In the block in which the data of the coefficient part is set to “0” by the coefficient operation unit 27 with respect to the data of the two-dimensional VLC code supplied from Determine the dimensional VLC code or DCT coefficients.

One way to determine the DCT coefficients is as follows:
The DCT coefficient obtained by setting the position of the maximum value of the two-dimensional VLC level or the maximum value of the absolute value of the DCT coefficient to “1” or “−1” and setting the rest to “0” for the DCT coefficient
Coefficient and, if necessary, the coefficient
Let it be the value of LC.

For a block having a plurality of maximum values at that time, the position where the code length becomes the shortest when encoded by two-dimensional VLC, and if they have the same code length, the zero run is the shortest. Is selected, and the coefficient at that position is set as a two-dimensional VLC value.

Further, another method for determining the DCT coefficient is as follows.
Regardless of the value of the two-dimensional VLC or DCT coefficient before conversion,
The type of the block, such as whether the coefficient block is a color difference signal for a luminance signal or a color difference signal for blue or red, and furthermore, the coefficient distribution of DCT coefficients before conversion has the least effect on the image quality of the converted image, and the two-dimensional VLC code. The position of the coefficient whose length becomes shorter is determined in advance, and the DCT coefficient after conversion is determined based on the determined position.

The converted DCT coefficients determined by the coefficient selection unit 24 according to these methods are supplied to the coefficient operation unit 27, and the coefficient operation unit 27 performs rate conversion of the supplied coefficients, and performs the rate-converted DCT. Two-dimensional VLC of coefficient
The code is supplied to the DCT coefficient combining unit 28.

In the DCT coefficient combining unit 28, the supplied two-dimensional VLC code and data other than the DCT coefficient part supplied from the DCT coefficient separating unit 26 are recombined. The combined signal is supplied to the combining unit 15, where the value of the transfer rate is combined with the updated header data, and the combined signal is combined with an encoded signal such as an audio signal other than an image in the re-MUX unit. The combined signal is provided as a stream output of the image coding rate converter 10b.

As described above, the stream output signal from the image coding rate converter 10b according to the second embodiment is supplied, but the quantization scale code is not changed in the stream output, and the sequence output signal is supplied. A signal in which only the bit rate information and the DCT coefficient portion recorded in the header are updated, and the image coding rate converter 10b that updates the bit stream while maintaining the same data structure can be easily configured. Is shown.

Next, a description will be given of a third embodiment of the method for converting the rate of encoded image data with reference to the drawings. Figure 1
FIG. 5 shows the configuration of an image coding rate conversion device according to the third embodiment.

The image coding rate conversion device 10c
This is a device for performing rate conversion by inverse quantization / requantization as in the first embodiment. Compared to the first embodiment, a device between the inverse quantization / requantization unit 22 and the variable length encoding unit 25 is provided. Is provided with the above-described coefficient operation unit 27.

In the structure according to the first embodiment, the code amount is controlled by the converted quantization scale supplied from the code amount control unit 23. However, the quantized scale code exists in the data before the conversion. Code amount control in the case where the number of macro blocks is small and the change position of the quantization scale code is insufficient in the code amount control after conversion is performed as follows.

That is, the control is performed by the inverse quantization / requantization unit 2
In the coefficient operation unit 27 next to the second, the DCT after the requantization is performed.
By reducing some of the coefficients to "0" or a value smaller than the original coefficient to reduce the coefficients, the two-dimensional VLC code length encoded by the subsequent variable-length encoding unit is reduced, The code amount after conversion in the code amount controller 23 is set to a required value.

The position of the DCT coefficient to be reduced at that time and the specific method of reducing the coefficient are supplied from the code amount controller 23. For example, when the two-dimensional VLC processing is performed, the DCT coefficient is located at the rear end. There is a method of reducing the coefficients in order from the coefficient to be performed.

In the third embodiment shown here, the coefficient operation unit 27 performs the same coefficient reduction as described above.

The blocks whose coefficients are reduced by the coefficient operation unit 27 or whose coefficients supplied from the inverse quantization / requantization unit 22 are all 0 are selected by the coefficient selection unit 24 described above. The coefficients to be left are supplied from the coefficient operation unit 27.

As described above, the stream signal whose coding rate has been changed by the image coding rate conversion apparatus 10c shown in the third embodiment is supplied. A rate conversion device that performs the same operation can be configured even when it is added to the image encoding rate conversion device 10a obtained by modifying the first embodiment shown in FIG.

As described above, the first and second methods for converting the transfer rate without converting the bit stream signal coded based on the MPEG-2 system to the pixel level and maintaining the same bit stream data structure. -The third embodiment has been described.

In the rate conversion method performed by these embodiments, the supplied moving image signal in units of frame images is divided into a plurality of image data for each pixel block of, for example, 16 pixels × 16 pixels. A motion vector amount is obtained for each pixel block, and while performing motion compensation prediction based on the obtained motion vector amount, orthogonal transformation is performed on image data of the pixel block by DCT transformation or the like,
Quantize the data obtained by the orthogonal transform, generate a coded image signal by performing variable-length coding on the data obtained by the quantization, and orthogonal transform related to the generation of the coded image signal, and Information on coding parameters such as motion prediction is obtained as coding information, a header signal including the obtained coding information is generated, and the generated coded image signal and header signal are combined to form a first signal. Obtained as a first encoded bit stream compressed and encoded according to the transfer rate, and the data of the obtained bit stream is supplied and is lower than the first transfer rate or the first.
A rate conversion method for image encoded data obtained by converting the signal into a bit stream signal encoded at a second transfer rate different from the transfer rate of Coded image signal and coding parameter information are separated, and based on the coded information obtained by separation, coefficient data for each pixel block obtained by orthogonally transforming the coded image signal is obtained, and the obtained image is obtained. Update image data is generated by deleting one or more of the plurality of quantized coefficients of the quantized data related to the data and deleting the other quantized coefficients. Since it is possible to generate the second encoded data at the second transfer rate by combining the encoded information whose transfer rate information and the like are updated based on the obtained encoded information, The delay time for the signal conversion can be reduced as compared with the conversion of the transfer rate by converting the converted signal into an image signal and then encoding by re-conversion, and temporarily storing the image signal for the signal conversion. , The conversion method is easy, and the generation of the updated coded data is such that the transfer rate is converted while maintaining the same data structure in the first coding method. Each converted macro block is converted to the same macro block type and CBP signal as before conversion, and when rate conversion is performed by requantization, the same as the updated macro block position of the quantization scale before conversion is used. Since the rate conversion is performed by updating the quantization scale only at the position of, only the DCT coefficient part is converted, D when performing the conversion
When only the CT coefficient part and the quantization scale code can be used, the macroblock type and CBP are not changed before and after the conversion, and the data operated by the conversion directly operates the variable length code of the DCT coefficient part. Can simplify the processing involved in rate conversion by only changing the DCT coefficient part and the quantization scale code in the rate conversion by requantization. The attached header signal can also be the minimum change of the transfer rate information and the like, and the generation of the header signal can be performed without waiting for all the image data corresponding to the header signal to be updated, thereby realizing the rate conversion. Although the method is easy and the configuration of the device for rate conversion can be simple, such an easy method and simple method can be used. Coded signal generated by such a device can change the transfer rate harmoniously in the entire bit stream, so even when trying to obtain a bit stream converted to a coded signal with a low transfer rate, the image quality can be improved. A signal converted to a coded signal with little deterioration can be obtained.

In the above-described embodiment, the encoding method of the bit stream to be converted is mainly MPEG-2. However, the encoding method is not limited to MPEG-2, and a predetermined amount of encoding is performed. An encoding method having a data structure similar to that of MPEG-2, in which information on the encoding parameter is attached to the image data as header information and a bit stream is attached to the image data;
For example, the present invention can be applied to a bit stream encoded by a method such as the MPEG-1 method, H261, and other MPEG-4, MPEG-7, and MPEG-21.

The code amount control before and after the rate conversion shown in these examples can be applied to either fixed bit rate control or variable bit rate control.

[0148]

According to the first aspect of the present invention, the transfer rate can be converted without converting the coded signal to the pixel level and maintaining the same data structure of the coded signal. The amount of data stored for signal processing is small, the delay time for signal processing is short, the image quality is low, and there is no need to make complicated code changes such as changing headers such as CBP. There is an effect that it is possible to provide a simple image encoded data rate conversion method that can easily perform control.

According to the second aspect of the invention, in particular, since the transfer rate of the coded signal is converted using the updated quantization scale, in addition to the effect of the first aspect, the image quality is further reduced. In addition, there is an effect that it is possible to provide a rate conversion method of image coded data in which code amount control can be easily performed.

According to the third aspect of the present invention, in particular, the deletion of image data relating to a pixel block is performed in order from the higher-order coefficient data obtained by orthogonally transforming the pixel block. In addition to the effect of item 1, there is an effect that it is possible to provide a rate conversion method of image encoded data in which image quality deterioration is further reduced and code amount control can be easily performed.

Further, according to the fourth aspect of the present invention, the image data is deleted while leaving the quantization coefficient giving the maximum absolute value particularly regarding the orthogonal transform coefficient. In addition to the effects of
In addition, there is an effect that it is possible to provide a method of converting the rate of encoded image data in which the code amount can be easily controlled.

Furthermore, according to the fifth aspect of the present invention, in particular, the variable length coding is performed so that the quantization coefficient giving the shortest code length is left so that the image data is deleted. In addition to the effect of item 1, there is an effect that it is possible to provide a method of converting the rate of encoded image data in which image quality deterioration is further reduced and code amount control can be easily performed.

Further, according to the invention of claim 6, image data can be reduced by substituting a coefficient value having a short code length of the two-dimensional VLC and having little effect on reproduction image quality, instead of a plurality of quantization coefficients. The transfer rate can be converted without converting the coded signal to the pixel level and maintaining the same data structure of the coded signal, so that the data storage amount for signal processing is reduced. Simple image coding that can easily control the code amount, with less delay time for signal processing, less deterioration in image quality, no need for complicated code changes involving header changes such as CBP There is an effect that a data rate conversion method can be provided.

According to the seventh aspect of the present invention, the transfer rate can be converted without converting the coded signal to the pixel level and maintaining the same data structure of the coded signal. , The amount of data stored for signal processing is small, the delay time for signal processing is short, the image quality is not deteriorated, and there is no need to make complicated code changes such as changing headers such as CBP. There is an effect that it is possible to provide a configuration of a simple image encoded data rate conversion device that can easily perform control.

[Brief description of the drawings]

FIG. 1 is a schematic block diagram of an image coding rate conversion device according to an embodiment of the present invention.

FIG. 2 is a schematic block diagram of an MPEG-2 encoding device.

FIG. 3 is a diagram illustrating a structure of a picture type related to motion compensation prediction in the MPEG-2 encoding method.

FIG. 4 is a diagram illustrating a structure of pixel data relating to transform coding of the MPEG-2 coding method.

FIG. 5 is a diagram showing a data structure of a bit stream transmitted by the MPEG-2 system.

FIG. 6 is a diagram showing a data structure in a macroblock layer encoded according to the MPEG-2 system.

FIG. 7 shows a macroblock type VLC table for an I picture encoded according to the MPEG-2 system.

FIG. 8 shows a macroblock type VLC table for a P picture coded according to the MPEG-2 system.

FIG. 9 shows a macroblock type VLC table for a B picture encoded according to the MPEG-2 system.

FIG. 10 is a diagram showing a data structure in a block layer encoded by the MPEG-2 system.

FIG. 11 is a diagram illustrating a configuration of an image coding rate conversion device according to a first embodiment of the present invention.

FIG. 12 is a diagram for explaining the existence of a quantized scale code before and after rate conversion according to the embodiment of the present invention.

FIG. 13 is a diagram showing a configuration of an image coding rate conversion device according to a modification of the first embodiment relating to the implementation of the present invention.

FIG. 14 is a diagram illustrating a configuration of an image coding rate conversion device according to a second embodiment of the present invention.

FIG. 15 is a diagram showing a configuration of an image coding rate conversion device according to a third embodiment of the present invention.

FIG. 16 is a diagram showing a configuration of a rate conversion device in a conventional example.

[Explanation of symbols]

10, 10a, 10b, 10c Image coding rate converter 11 DEMUX unit 12 Input code amount counter 13 Header separation unit 14 Header data storage unit 15 Header combining unit 16 Output code amount counter 17 Re-MUX unit 20, 20a, 20b, 20c Image rate conversion unit 21 Variable length decoding unit 22 Inverse quantization / requantization unit 22a Inverse quantization unit 22b Requantization unit 23 Code amount control unit 24 Coefficient selection unit 25 Variable length encoding unit 26 DCT coefficient separation unit 27 Coefficient Operation unit 28 DCT coefficient combining unit 60 MPEG-2 encoder 61 Subtractor 62 DCT unit 63 Quantizer 64 Code amount controller 65 Variable length encoder 66 Buffer 71 Inverse quantizer 72 IDCT unit 73 Adder 74 Frame memory 75 Motion compensation predictor 81 data separation circuit 82 inverse VLC circuit 83 inverse quantum Digitizer 84 bit rate control circuit 85 quantizer 86 VLC circuit 87 coupling circuit 88 buffer circuit

 ──────────────────────────────────────────────────続 き Continued on the front page F term (reference)

Claims (7)

[Claims] 1. A moving image signal is divided into a plurality of image data for each pixel block of a predetermined size, a motion vector amount is obtained for each of the divided pixel blocks, and a motion vector amount is calculated based on the obtained motion vector amount. Compensation prediction, orthogonal transformation of the image data of the pixel block, quantization of the data obtained by the orthogonal transformation to obtain quantized data, and variable-length encoding of the obtained quantized data. Generates an encoded image signal, obtains information on encoding parameters related to the generation of the encoded image signal as encoded information, generates a header signal including the obtained encoded information, and generates them. The encoded image signal and the header signal are combined to obtain first encoded data compressed and encoded at a first transfer rate, and the obtained first encoded data is supplied. A rate conversion method of image encoded data obtained by converting the encoded data of (a) into second encoded data compressed and encoded at a second transfer rate, wherein the encoding is performed based on the first encoded data. A first step of obtaining an image signal and the coded information, and based on the coded information obtained in the first step,
Obtain image data for each pixel block obtained by orthogonally transforming the coded image signal, and leave one or more of a plurality of quantized coefficients of quantized data relating to the obtained image data and perform other quantization. A second step of generating updated image data by deleting a coefficient; an updated image data generated by the second step; and an updated code based on the encoded information obtained in the first step. A third step of combining the coding information to obtain the second coded data. 2. The method according to claim 2, wherein the updated image data in the second step is a method for storing at least one of a plurality of quantized coefficients obtained by requantizing the image data with the updated quantization scale. 2. The method according to claim 1, wherein the conversion coefficient is generated by deleting the conversion coefficient. 3. The updated image data in the second step is sequentially deleted from a plurality of quantized coefficients of the quantized data related to the image data, from a quantized coefficient related to a higher-order orthogonal transform coefficient. 2. The method according to claim 1, wherein the rate conversion is performed. 4. The updated image data in the second step is a quantized coefficient that gives a maximum absolute value with respect to an orthogonal transform coefficient among a plurality of quantized coefficients of quantized data relating to the image data. 2. The method according to claim 1, wherein the image data is generated so as to leave the data. 5. The update image data in the second step, wherein, among a plurality of quantized coefficients of the quantized data relating to the image data, a quantized coefficient for performing the variable length encoding to provide a shortest code length. 2. The method according to claim 1, wherein the image data is generated so as to leave the conversion coefficient. 6. A moving image signal is divided into a plurality of image data for each pixel block of a predetermined size, a motion vector amount is obtained for each of the divided pixel blocks, and a motion vector amount is calculated based on the obtained motion vector amount. Compensation prediction, orthogonal transformation of the image data of the pixel block, quantization of the data obtained by the orthogonal transformation to obtain quantized data, and variable-length encoding of the obtained quantized data. Generates an encoded image signal, obtains information on encoding parameters related to the generation of the encoded image signal as encoded information, generates a header signal including the obtained encoded information, and generates them. The encoded image signal and the header signal are combined to obtain first encoded data compressed and encoded at a first transfer rate, and the obtained first encoded data is supplied. A rate conversion method for image-encoded data obtained by converting encoded data of (a) into second encoded data that is compression-encoded at a second transfer rate, wherein the encoding is performed based on the first encoded data. A first step of obtaining an image signal and the coded information, and based on the coded information obtained in the first step,
Image data for each pixel block obtained by orthogonally transforming the coded image signal is obtained. Instead of a plurality of quantized coefficients of quantized data related to the obtained image data, a two-dimensional V
A second step of generating updated image data by using a coefficient value having a short LC code length and having little effect on image quality; updated image data generated by the second step; and the first step A third step of obtaining updated second encoded data by combining updated encoded information based on the encoded information obtained in (3). 7. A moving picture signal is divided into a plurality of image data for each pixel block of a predetermined size, a motion vector amount is obtained for each of the divided pixel blocks, and a motion vector amount is calculated based on the obtained motion vector amount. Compensation prediction, orthogonal transformation of the image data of the pixel block, quantization of the data obtained by the orthogonal transformation to obtain quantized data, and variable-length encoding of the obtained quantized data. Generates an encoded image signal, obtains information on encoding parameters related to the generation of the encoded image signal as encoded information, generates a header signal including the obtained encoded information, and generates them. The encoded image signal and the header signal are combined to obtain first encoded data compressed and encoded at a first transfer rate, and the obtained first encoded data is supplied. An image coding rate conversion device obtained by converting the coded data of (a) to (b) into second coded data compressed and coded at a second transfer rate, wherein the coded image signal is obtained from the first coded data. And a signal separating unit that obtains the encoded information; and obtains image data for each pixel block obtained by orthogonally transforming the encoded image signal, and obtains a plurality of quantized data of the quantized data related to the obtained image data. Image data converting means for generating updated image data by deleting one or more of the quantized coefficients and deleting other quantized coefficients; updating image data generated by the image data converting means; An image coding rate conversion apparatus, comprising: signal combining means for combining updated coding information based on the obtained coding information to obtain the second coded data.