JP2000050263A - Image coder, decoder and image-pickup device using them - Google Patents

Abstract

PROBLEM TO BE SOLVED: To use common circuits to the utmost for duplicate processing in the case of simultaneously realizing coding/decoding of an MPEG stream and a JPEG stream and to use a coding/decoding means through hardware and a coding/decoding means by software in common. SOLUTION: Common sharing of a quantization circuit 5 and an inverse quantization circuit 6 is devised for MPEG and JPEG processing. This is realized by providing two planes of memories storing a quantization matrix, Intra.Non- Intra coefficients are used for the MPEG and coefficients for luminance and color difference signals are used for the JPEG as the coefficients stored in them to cope with the respective processing. In the JPEG processing, a coding table is selected freely, depending on the image by conducting variable length coding 13/decoding 27 using the variable coding table, even for the software in addition to variable length coding 10/decoding 21 which correspond to a fixed coding table by the hardware and the compression efficiency is improved.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a data compression technique using variable length coding and decoding, and more particularly to an image coding and decoding apparatus for rapidly compressing and decompressing moving images and still images. It is.

[0002]

2. Description of the Related Art Moving picture coding and decoding methods include:
Currently MPEG (Moving Picture Expe
(rts Group) standard has been put to practical use as an international standard. Of the MPEG standards, MPEG-1 mainly for storage media is being commercialized as a standard for storing and transmitting moving images on personal computers and networks. JPEG (Joint Photog) is an international standard for compression and decompression of still images.
There is a standard called "Radical Experts Group", which is currently used as a standard in various media.

On the other hand, MPEs with a view to broadcasting, communications, etc.
The G-2 standard has been put to practical use, and an LSI (CODEC) integrating an encoder and a decoder has been announced.

[0004] The image compression technique of the MPEG system is a combination of frequency conversion using orthogonal transformation and variable-length coding. The MPEG2 standard is ISO / IEC1381.
8 is described.

[0005] The orthogonal transform referred to here is called DCT (Discrete Cosine Transform), and directly converts a block-divided image into a frequency domain. Next, after individually weighting (quantizing) the frequency components, the coefficient value 0
Are subjected to variable-length coding using a combination of the number of consecutive. In MPEG, compression processing is performed using this and motion-compensated inter-frame prediction for reducing redundancy in the time direction. On the other hand, the compression method of the JPEG method is M
It is the basis of the PEG standard and is based on ISO / IEC10
918. This also uses DCT transform and variable length coding, and the coding and decoding method is MP
It is almost common to EG.

[0006]

As media that will be important in the future, such as a camera that records and reproduces an MPEG stream, a still image (JPEG stream) is simultaneously recorded and reproduced because of its convenience and convenience. You need to be able to do it. However, since the camera itself is required to be lightweight, compact, and low power, encoding and decoding of MPEG streams and encoding of JPEG streams,
For decoding, it is desirable to use a common circuit for a portion where processing overlaps, but the above-described prior art is not mentioned as to a means for realizing the same. A first object of the present invention is to solve this problem and to realize an LSI that performs compression / expansion processing of moving images and still images.

Further, in the conventional JPEG processing, since the variable length encoding / decoding is performed using only the variable length encoding / decoding portion corresponding to the fixed encoding table provided in the LSI, the compression efficiency is constant. No further improvement was possible, and there was no freedom in processing. A second object of the present invention is to solve this problem.

[0008]

Means for Solving the Problems To achieve the first object of the present invention, the following means are adopted.

When the MPEG coding and the JPEG coding are compared, a part for converting an input image into blocks, a part for performing DCT conversion on the block-converted image data,
The part that quantizes the DCT-transformed one is common.
In the decoding, a part for inversely quantizing the variable-length decoded frequency data, a part for inverse DCT transform of the inversely quantized data, and a part for rearranging the inverse DCT transformed data into frame-based image data And are common. Therefore, when the moving image and the still image are compressed and decompressed by an LSI or the like, by sharing the circuit of the processing unit in the MPEG and JPEG processing, the circuit size of the LSI can be reduced and the power consumption can be reduced. In the present invention, the quantization / inverse quantization circuit among the above circuits to be shared in the processing of MPEG and JPEG is devised. The LSI has memories for storing quantization matrices for two planes, and the coefficients stored therein are subjected to intra-frame encoding (In
tra) inter-frame predictive coding (Non-Intr)
a), and in JPEG, for luminance and color difference signals,
MPE is stored in the memory for storing the quantization matrix.
In G, the coding for intra-frame coding (Intra) and inter-frame prediction coding (Non-Intra) is performed.
And JPEG. With this configuration, the MPEG processing and the JPEG processing can be performed by one quantization circuit.

In order to write the coefficients into the memory for storing the quantization matrix, a method of writing the coefficients from an external CPU is conceivable. Is very stressful. Generally, the external CPU needs to perform not only the encoding means but also other processing. However, when the external CPU becomes very heavy during the MPEG processing, the processing becomes heavy, and the operation becomes heavy. May slow down or stop working. On the other hand, since the fixed quantization coefficient is used in the MPEG processing, the quantization coefficient used in the MPEG can be stored in the ROM in the LSI. The configuration in which the quantization coefficient used in MPEG is written from the ROM in the LSI to the memory for storing the quantization matrix can prevent the load on the external CPU.

Further, the usability can be improved by writing the quantization coefficient used in MPEG into the memory for storing the quantization matrix when the power is turned on or when the MPEG mode is selected.

In MPEG processing, inverse quantization is required at the time of encoding in order to perform predictive encoding. If only one memory for storing the quantization matrix is used to perform the inverse quantization, it is necessary to perform the quantization and the inverse quantization alternately at different times during the MPEG processing. There is a problem that the processing speed of the quantization circuit must be increased in order to decrease the processing speed or to improve the processing speed. Therefore, by separately providing two memories, one for storing the quantization matrix and the other for storing the inverse quantization matrix, it is possible to perform both quantization and inverse quantization simultaneously. The speed of the MPEG processing can be improved without increasing the processing speed capability of the quantization circuit.

In order to achieve the second object of the present invention,
The processing to be performed by hardware such as an LSI is performed by software processing using a microcomputer. With this configuration, the degree of freedom in processing can be improved. In the JPEG processing, apart from the variable length coding / decoding part corresponding to the fixed coding table provided in the LSI, the software can perform variable length coding / decoding using an arbitrary coding table. The encoding table and the like can be freely selected according to the image, and the compression efficiency can be improved.

As described above, according to the moving picture (MPEG) and still picture (JPEG) encoding / decoding apparatus according to the present invention, the processing is implemented in an LSI by sharing a part of the circuit between the moving picture processing and the still picture processing. In such a case, the circuit scale can be reduced. Further, by performing a part of the processing with software external to the LSI, the degree of freedom of the processing is improved, and image compression can be performed efficiently.

[0015]

Embodiments of the present invention will be described below in detail with reference to the drawings.

FIG. 1 shows an MP according to a first embodiment of the present invention.
FIG. 1 is a block diagram of an encoding device that performs moving image encoding using an EG method and still image encoding using a JPEG method by sharing circuits.

In FIG. 1, reference numeral 1 denotes a block converter for converting input image data into a block of 8 × 8 pixels;
Is a subtractor that calculates the difference between the prediction data of the previous frame and the data of the current frame, and 3 is the inter-frame prediction (Non
A selector for switching processing between (Intra) coding, intra-frame (Intra) coding, and JPEG coding;
Is a DCT calculator for DCT-transforming the input image data in block units, 5 is a quantizer for quantizing the DCT-transformed data by a predetermined coefficient, and 6 is an inverse quantization for returning the quantized data to the original. , 7 is the ID of the dequantized data
IDCT calculator for performing CT conversion and returning to original image data, 8
Is an adder that adds the result of the IDCT transform to the image data of the previous frame, 9 is a prediction memory for storing the current image as a prediction image of the next frame, and 10 is default data that has been quantized in JPEG encoding. A variable-length encoder for performing variable-length encoding using a fixed-value) encoding table; 11 denotes a fixed-value encoding table (fixed table) that quantizes data quantized in MPEG encoding according to the MPEG standard A variable-length encoder for performing variable-length encoding by using H.264, encoding means for performing hardware processing at high speed for processings 1 to 11; and 13, arbitrary encoding of data quantized by JPEG encoding Table (variable table)
, A selector for selecting an arbitrary result from 10 encoding results and 13 encoding results in JPEG encoding, and 15 a MPE.
A selector for switching the processing between G encoding and JPEG encoding,
Reference numeral 16 denotes a stream multiplexer for adding a header or the like necessary for creating an MPEG / JPEG stream to encoded data, 17 a buffer for temporarily holding the generated stream before outputting it to the outside, and 18 a buffer for 13 to 17. Encoding means for performing processing mainly by software are shown.

In the case of the encoding of the MPEG system, the operation of each processing is as follows.

The input image data is converted into a block
Is converted to block data. The data of the previous frame is subtracted by the subtracter 2 from the converted block data. In the inter-frame prediction coding mode, the selector 3
Is selected and sent to the DCT calculator 4. At the time of intra-frame encoding, the selector 3 selects the output of the blocker 1. The DCT calculator 4 performs DCT conversion on the input block data. The DCT-transformed data is quantized by a quantizer 5 using a predetermined coefficient. The quantized data is sent to the variable-length encoder 11, and variable-length encoding using a coding table defined by the MPEG standard is performed using a combination of the number of consecutive 0 coefficients and non-zero coefficients as an event. Will be The selector 15 selects the output of the variable length encoder 11 and sends it to the stream multiplexer 16. In the stream multiplexer 16, the variable-length encoded data is
Various kinds of information (a header, a quantization table, a quantization parameter, a motion vector, and the like) necessary as an MPEG stream are added, and the information is sent to the buffer 17. The buffer 17 outputs the encoded MPEG stream to the outside at a predetermined rate. In the inter-frame prediction coding, inverse quantization is performed on the quantized data by the inverse quantizer 6. The result is sent to the IDCT converter 7 and returned to the original image data, then added to the data of the previous frame by the adder 8 and recorded in the prediction memory 9 for predictive coding of the next frame.
At the time of predictive encoding, data is read from the predictive memory 9 and subtracted by the subtracter 2 from the data of the current frame.
The subtracted data is sent to the selector 3.

In the case of JPEG encoding, the operation of each process is as follows.

The input image data is converted into a block
Is converted to block data. In the case of JPEG processing,
In the selector 3, the output of the blocker 1 is always selected.
It is sent to the CT calculator 4. The DCT calculator 4 performs DCT on the input block data. The DCT-transformed data is quantized by a quantizer 5 using a predetermined coefficient. The quantized data is transmitted to the variable length encoders 10 and 1
The variable length coding is performed using the combination of the number of consecutive 0 coefficients and the non-zero coefficient as an event. At this time, JP
Either the variable length encoder 10 using hardware using a fixed table recommended by the EG standard or the variable length encoder 13 using software using any other variable table selectively operates The output that is operating is selected by the selector 14, further selected by the selector 15, and sent to the stream multiplexer 16. For this selection, the user previously determines whether to use a fixed table or a variable table for variable-length coding, and the mode is switched to determine which to use. In a special image, the event may be a combination of the number of consecutive 0 coefficients and a non-zero coefficient. In some cases, the compression ratio can be improved by using another table rather than by using a fixed table. Therefore, when the combination of the number of consecutive 0 coefficients and the non-zero coefficient is special,
Variable length encoder 10 (fixed) according to image content and characteristics
Or the output of the variable-length encoder 13 (variable) may be automatically switched.
The compression efficiency can be improved by selecting which one has higher encoding efficiency. Specifically, the total data amount of the output data of the variable-length encoder 10 (fixed) and the output data of the variable-length encoder 13 (variable) are compared, and the smaller the total data amount, the higher the compression ratio. Select the output with the smaller total amount. The variable table used in the variable length encoder 13 may be one in which some types are incorporated in a program in advance. If an interface is provided in the encoding means 18 so that a user or the like can rewrite the variable table of the variable-length encoder 13, the degree of freedom of encoding can be further improved, and the optimal compression ratio can be improved. Can be used to perform the sign. The stream multiplexer 16 adds various information (header, quantization table, quantization parameter, etc.) necessary as a JPEG stream to the variable-length coded data, and performs variable-length decoding after adding the various information. Is output to the outside.

As described above, according to the first embodiment, since the same blocker 1, DCT transformer 4, and quantizer 5 can be used for JPEG encoding and MPEG encoding,
The hardware circuit of the PEG / JPEG encoder can be reduced. At this time, the variable length encoders 10 and 11 can share a circuit between the MPEG processing and the JPEG processing by switching the fixed table.

In the first embodiment, the output of the variable length encoder 10 and the output of the variable length encoder 11 are selected by the selector 15, but the output of the variable length encoder 10 and the output of the variable length encoder 11 are selected. A selector may be provided in the encoding means 12 to output a selection output from the encoding means 12, and an output of the encoding means 12 and the variable length encoder 13 may be selected by an external selector.

In the first embodiment, the output of the variable length encoder 10 and the output of the variable length encoder 13 are selected by the selector 14. However, the output of the variable length encoder 10 and the output of the quantization circuit are selected. May be provided in the encoding means 12, and a selected output from the encoding means 12 may be output to the encoding means 18.

Next, the period required for processing related to the image format in the MPEG / JPEG processing will be described. FIG. 5 is an explanatory diagram of an image format in the MPEG / JPEG standard. In FIG.
(C) is the data rate of the luminance signal and the color difference signal,
(B) and (d) show the ratio of the sampled pixels of the luminance signal and the chrominance signal in one frame.

According to the MPEG2 standard, the number of pixels to be subjected to the encoding / decoding processing is a standard subset MP ＠ M
L (Main Profile at Main Le)
vel), 720 horizontal pixels × 480 vertical pixels.
This image format is called [4: 2: 0], and the sampling density of the chrominance signal is 1/4 that of the luminance signal as shown in (a) and (b). According to the JPEG standard, in addition to the [4: 2: 0] format which is the same as the MP @ ML of MPEG2, the number of pixels of the color difference signals (Cb, Cr) as shown in (c) and (d) is twice as large in the vertical direction. [4: 2: 2] format exists. There is no [4: 2: 2] format in MP @ ML. MPEG
In JPEG encoding / decoding, each frame is divided into macroblocks (16 pixels × 16 pixels), and processing is performed in fixed time slots in macroblock units. Since either format can be used in JPEG, the JPEG image format is [4:
2: 0], if JPEG processing is performed using an MPEG2 processing circuit, processing of one macroblock of a JPEG image can be performed in the same period as processing of one macroblock of an MPEG image. Become. Therefore, by this, motion JPEG (moving image is JPE every frame)
G encoding / decoding). When the JPEG image is in the [4: 2: 2] format, the processing for one macroblock of the JPEG image is performed over two macroblock periods of the MPEG image because the number of pixels is large. FIG. 6 is a diagram for explaining a period and an image format required for MPEG and JPEG encoding / decoding processing for one macroblock. When the JPEG encoding / decoding processing is performed using the MPEG encoding / decoding processing circuit, the JPEG image format is [4: 2:
0], the period required for encoding / decoding one macroblock is the same as the one macroblock processing period of MPEG (see (a) and (b)). So at this time,
Images of one JPEG frame can be processed in real time. As a result, in the case of NTSC, JPEG encoding / decoding of 30 frames per second becomes possible, and motion JP
EG can be realized. When the JPEG image format is [4: 2: 2], the processing period for one macroblock exceeds the processing period for one macroblock of MPEG, and therefore, as shown in FIG. The processing for one JPEG macroblock is performed over the time of the block period. Although FIG. 6 shows a state in which only one-third of the operation is performed during one macroblock process and a one-third pause is performed, the ratio of the operation and pause period can be changed.

FIG. 2 shows a second embodiment of the present invention.
FIG. 3 is a block diagram of a decoding device that performs video decoding using the EG method and still image decoding using the JPEG method by sharing circuits.

In the figure, reference numeral 19 denotes a buffer for temporarily storing a stream inputted at the time of MPEG decoding, 51 a stream separator for reading additional information necessary for decoding from the inputted stream, and 20 a buffer for reading from the buffer 19. A variable-length decoder for decoding the MPEG stream obtained by using an encoding table (fixed table) defined by the MPEG standard, 21 is a variable-length decoder for decoding by using a default fixed table at the time of JPEG decoding , 27
Is a variable length decoder for decoding using an arbitrary encoding table (variable table) other than the fixed table at the time of JPEG decoding, and 28 is a quantization table for image compression at the time of image compression, a quantization parameter, etc. Decoding means for performing the processing of separating data 51 mainly by software;
Reference numeral 22 denotes a selector which is used at the time of JPEG decoding and selects a decoding result in accordance with an encoding table, 23 denotes a selector which switches signals to be decoded by MPEG decoding processing and JPEG decoding processing, and 24 denotes MPEG intra-frame decoding. , MPEG
A selector for switching processing between inter-frame prediction decoding and JPEG decoding, 25 is a format converter for performing image size conversion as needed, 26 is 20 to 25, and 6 to 9
And decoding means for performing the processing at high speed by hardware. Other blocks are the same as those in the first embodiment shown in FIG.

In the case of MPEG decoding, the operation of each process is as follows.

The input MPEG stream is stored in the buffer 19 and then input to the variable length decoder 20. In the selector 23, the result of the variable length decoder 20 is selected and sent to the inverse quantizer 6. The inverse quantizer 6 performs inverse quantization using the quantization table and the quantization parameter included in the stream and extracted by the stream separator 51. The inversely quantized data is subjected to IDCT conversion by the IDCT calculator 7 and returned to image data in block units, and then added to the data of the prediction frame stored in the prediction memory 9 by the adder 8 to be converted by the format converter 25. Sent to The format converter 25 converts the image size and outputs it to the outside.

In the case of JPEG decoding, the operation of each process is as follows.

The input stream is obtained by extracting the header information in the stream by the stream separator 51 and reading out the encoding table used for encoding, and then reading it out from the fixed table recommended by the JPEG standard. The variable-length decoder 21 or 27 operates selectively depending on whether it is or not. The decoder 21 performs decoding using the fixed table, and the decoder 27 performs decoding using the variable table. The selector 22 selects 21 decoding results for the fixed table and 27 decoding results for the other variable tables according to the encoding table used.
Is output to In the case of JPEG decoding, the output of the selector 22 is selected by the selector 23 and sent to the inverse quantizer 6. The inverse quantizer 6 performs inverse quantization using the quantization table and the quantization parameter included in the stream. The inversely quantized data is processed by
The image data is subjected to CT conversion, returned to image data in block units, sent to a format converter 25 through a selector 24, and output. As to which of the output of the variable-length composite 27 and the output of the variable-length composite 21 is to be selected, a configuration may be employed in which the user switches between modes to determine which to select. The selector 22 determines that the encoding is performed using the variable table, and selects the output of the variable-length decoding unit 21 (fixed) or selects the output of the variable-length decoding unit 27.
It may be configured to switch whether to select (variable) output. In the case of JPEG, even if it is fixed, it is necessary to put an encoding table in the stream. Therefore, it is sufficient to determine whether to process the fixed value or the variable value by looking at the encoding table in the stream. In addition,
The variable table used for the variable-length decoding 13 may be a table in which several types are incorporated in the program in advance, or an interface is provided in the encoding means 18 so that the user or the like can use the variable-length decoding. If the variable table of the decoder 27 is configured to be rewritable, the degree of freedom of encoding can be further improved, and encoding can be performed with an optimal compression ratio.

As described above, according to the second embodiment, the inverse quantizer 6 and the IDCT are used for JPEG decoding and MPEG decoding.
The arithmetic unit 7 can be shared, and the hardware circuit of the MPEG / JPEG decoder can be reduced. Also, the variable length decoder 2
0 and 21 are MPEG tables by switching the fixed table.
Circuits can be shared between processing and JPEG processing.

In the second embodiment, the variable length decoders 20 and 21 are separately input.
May be shared and distributed to the variable length decoders 20 and 21 in the decoding means 26.

FIG. 3 shows a third embodiment of the present invention, JP.
FIG. 2 is a block diagram of a still image encoding device using the EG method. In the figure, reference numeral 29 denotes hardware encoding means for performing high-speed JPEG compression processing, and reference numeral 30 denotes encoding means for mainly performing processing by software. The blocks denoted by other reference numerals are the same as those in FIG.

The input image data is converted into data of 8 pixels × 8 pixels in a block unit by the block converter 1. The converted data is sent to the DCT calculator 4 where it is DCT-converted in block units. The DCT-transformed data is quantized by a quantizer 5 using a predetermined coefficient.
The quantized data is sent to the variable-length encoder 10, and variable-length encoding using a fixed table is performed using a combination of the number of consecutive 0 coefficients and a non-zero coefficient as an event. The quantization result is sent to the variable-length encoder 13 at the same time, and variable-length encoding is performed using another variable table adjusted according to the characteristics of the image and the like. The selector 14 selects one of the coding result using the fixed table and the coding result using the variable table, which has higher coding efficiency, according to the characteristics of the image to be compressed, and sends the selected result to the stream multiplexer 16. Specifically, the compression rate can be increased by comparing the total amount of output data of the variable length encoder 10 and the variable length encoder 13 and selecting the smaller amount of data. The user may determine which of the selectors 14 to select, and the setting may be made from outside the encoding unit 30. The stream multiplexer 16 adds various information (a header, a quantization table, a quantization parameter, an image size, etc.) necessary for a JPEG stream to the variable-length-encoded data, and outputs the data.

FIG. 4 shows a fourth embodiment of the present invention.
It is a block diagram of the still image decoding device using PEG system. In the figure, reference numeral 51 denotes a stream separator for extracting information necessary for decoding from an input stream; 27, a variable-length decoder for decoding a stream compressed using an arbitrary encoding table (variable table); 52 is 5
Decoding means for performing processing mainly using software, including 1, 27, etc .; 21, a variable length decoder for performing decoding using a fixed table recommended by the JPEG standard; Processing 21, 22, 6, 7, 2
5 represents decoding means for performing high-speed decoding using hardware. The other blocks are the same as those in FIG.

From the input stream, the encoding table in the header is extracted by the stream separator 51 and sent to the variable length decoders 21 and 27. The decoder 21 selectively performs decoding using the fixed table, and the decoder 27 selectively performs decoding using the variable table. Selector 2
In 2, a decoding result is selected according to the encoding table in the stream and sent to the inverse quantizer 6. Note that a configuration may be adopted in which the user determines which of the selectors 22 is to be selected, and which is selected from outside the encoding unit 29. In the inverse quantizer 6, inverse quantization is performed using the quantization table and the quantization parameter included in the stream. The inversely quantized data is subjected to IDCT conversion by the IDCT calculator 7 and is returned to image data in block units, and is then rearranged into frame images by the format converter 25 and output.

FIGS. 7, 8 and 9 show MPEG / JPE
FIG. 3 is an explanatory diagram of a quantizer / inverse quantizer in G processing.

When a common processing circuit is used for MPEG encoding / decoding and JPEG encoding / decoding, almost the same circuit can be used for image blocking, DCT transformation, IDCT transformation, and framing. However, the quantization tables used in MPEG and JPEG are different for quantization and inverse quantization. MPEG coding methods include a case where compression is performed by intra-frame coding (Intra coding) and a case where compression is performed using a difference between frames (Non-Intra coding).
Encoding). Quantization in MPEG and JPEG is generally performed according to the following equation.

Q = {D (i, j) × 16} / {Qs × W (i, j)} (i, j = 0 to 7) (Equation 1) where D (i, j) Is an input DCT result, Qs is a parameter for controlling the generated code amount, W (i, j) is a quantization table value corresponding to the coordinate position (i, j) of the DCT result, and Q is a quantization result. Respectively. A circuit for realizing this is as shown in FIG. The coordinate position at this time is shown in FIG.

The inverse quantization in MPEG and JPEG is generally performed according to the following equation.

IQ = D ′ (i, j) × {Qs × W (i, j)} / 16 (i, j = 0 to 7) (Equation 2) where D ′ (i, j) Is the input variable-length decoding result,
Qs is a parameter for controlling the generated code amount, W
(I, j) represents a quantization table value corresponding to the coordinate position (i, j) of the variable length decoding result, and IQ represents an inverse quantization result. FIG. 8 shows a circuit for realizing this. The coordinate position at this time is shown in FIG.

In the MPEG standard, Intra encoding and N
A different quantization table is used in on-Intra coding. As shown in FIG. 9, the quantization table (default value in the MPEG standard) has 8 × 8 elements corresponding to each pixel in the block. Each value of this quantization table can be selected according to the image.
At the time of inverse quantization, an arbitrary table value (2 blocks: 12
8 elements) into a storage element such as a memory, and Int
They are switched and used adaptively in ra encoding and Non-Intra encoding. According to the JPEG standard, different quantization tables are used for the quantization and inverse quantization of the luminance signal and the quantization and inverse quantization of the chrominance signal. Therefore, a quantization table corresponding to the luminance signal and a quantization table corresponding to the chrominance signal are used. Are written in the memory in the same manner as in the MPEG processing, and are switched and used in processing the luminance signal and the color difference signal. As a result, the quantization / inverse quantization circuit can be shared between MPEG and JPEG.

FIG. 10 shows a fifth embodiment of the present invention.
MPEG / JPEG used in the first and second embodiments
FIG. 2 is a block diagram in which a common quantizer 5 and an inverse quantizer 6 are expanded.

In the figure, 31 is a memory for storing all 128 element values of the quantization table, and 32 is a memory 31
, A quantization table value read out from the quantizer 5, a DCT coefficient input to the quantizer 5, a quantization result output from the quantizer 5, a DCT coefficient output from the inverse quantizer 6, 36 is a variable length decoding result input to the inverse quantizer 6, 37 is an address signal for reading data from the memory, 38 is a selector for selecting reading of addresses 0 to 63 and reading of addresses 64 to 127, 39
Is an address generator for generating addresses 0 to 63, 40 is an address generator for generating addresses 64-127, 41
Is a switching signal of the selector 38, 42 is MPEG processing and J
Selector for controlling switching of addresses in PEG processing, 5
0 is the value stored in the quantization table from the external CPU
1 represents an input signal to be written into “1”.

In the MPEG quantization / inverse quantization, when a default quantization table is used, the quantization table shown in FIG.
Write to. At this time, addresses 0 to 63 of the memory 31
(Area A) has an Intra quantization table, address 6
A Non-Intra quantization table is written in 4-127 (area B). However, there is no problem even if the writing area is reversed. In the MPEG processing, the selector 42 selects In
The switching signal of tra and Non-Intra is selected and becomes the switching signal of the selector 38. At the time of Intra encoding / decoding, the selector 38 inputs the address signal generated by the address generator 39 to the memory 31 and reads the quantization table from the area A. At this time, a quantization table corresponding to the coordinate position (i, j) of the DCT coefficient 33 is input to the quantizer 5, and the coordinate position (i, j) of the variable-length decoding result 36 is input to the inverse quantizer 6. Is input. At the time of Non-Intra encoding / decoding, the selector 38 inputs the address signal generated by the address generator 40 to the memory 31 and reads the quantization table from the area B. At this time, the DCT coefficient 3
The quantization table corresponding to the coordinate position (i, j) of the variable length decoding result 36 is input to the inverse quantizer 6. . Note that the above (i, j) corresponds to those shown in FIGS. 7B and 8B.

In the JPEG processing, the quantization table for the luminance signal is set in the area A, and the quantization table for the color difference signal is set in the area B.
Are written by the input 50 from the external CPU.
At the time of quantization / inverse quantization, a selector 42 selects a switching signal for a luminance signal and a color difference signal, which becomes a switching signal for the selector 38. The selector 3 is used for encoding / decoding the luminance signal.
At 8, the address signal generated by the address generator 39 is input to the memory 31, and the quantization table is read from the area A. At this time, a quantization table corresponding to the coordinate position (i, j) of the DCT coefficient 33 is input to the quantizer 5, and the coordinate position (i, j) of the variable-length decoding result 36 is input to the inverse quantizer 6.
A quantization table corresponding to j) is input. When encoding / decoding the color difference signal, the selector 38 controls the address generator 4
The address signal generated by 0 is input to the memory 31, and the quantization table is read from the area B. At this time, a quantization table corresponding to the coordinate position (i, j) of the DCT coefficient 33 is input to the quantizer 5, and the coordinate position (i, j) of the variable-length decoding result 36 is input to the inverse quantizer 6. Is input. Note that the above (i, j) corresponds to FIG.
8 (b).

In the fifth embodiment, the quantization table input from the external CPU is written. However, if the configuration is written from the outside, a load is imposed on the external CPU. MP
Since the quantization table used in the EG processing is a fixed value, a default quantization table used in the MPEG processing is stored in the internal ROM of the encoding unit 10 and is loaded from the internal ROM to the memory 31. If, CP
U can be prevented from being overloaded.

This loading is automatically performed from the internal ROM to the memory 31 according to the processing mode, such as at the time of resetting or switching between the respective modes of the MPEG processing and the JPEG processing. With this configuration, usability can be improved and it is convenient.

FIG. 11 shows a sixth embodiment of the present invention.
MPEG JPE used in the first and second embodiments
FIG. 3 is a block diagram in which a quantizer 5 and an inverse quantizer 6 that can perform independent quantization and inverse quantization processes for G sharing are expanded.

In the figure, reference numeral 31 denotes a memory for storing all 128 element values of a quantization table for quantization;
2 is a quantization table value sequentially read from the memory 31, 33 is a DCT coefficient input to the quantizer 5, 34 is a quantization result output from the quantizer 5, and 37 is an address for reading data from the memory. A signal 38, a selector for selecting the reading of addresses 0 to 63 and a reading of addresses 64 to 127; 39, an address generator for generating addresses 0 to 63; 40, an address generator for generating addresses 64 to 127; Is a switching signal of the selector 38, 42 is a selector for controlling the switching of the addresses by the MPEG processing and the JPEG processing, 50 is an input signal for writing the element value of the quantization table from the external CPU to the memory 31, and 35 is the inverse quantization signal. DCT coefficient output from the transformer 6
6 is a variable length decoding result input to the inverse quantizer 6;
Is a memory for storing all 128 element values of the quantization table for inverse quantization, 53 is a quantization table sequentially read from the memory 43, 44 is an address signal for reading data from the memory 43, and 45 is an inverse signal. A selector for selecting the reading of addresses 0 to 63 and the reading of addresses 64 to 127 during quantization, 46 is an address generator for generating addresses 0 to 63, 47 is an address generator for generating addresses 64 to 127, 48 Represents a switching signal of the selector 45, 49 represents a selector for controlling the switching of addresses by MPEG processing and JPEG processing, and 50 represents an input signal for writing an element value of the quantization table from the external CPU to the memory 43.

In the case of using the default quantization table in the MPEG quantization / inverse quantization, the quantization table shown in FIG.
And writing to the memory 43 at the same time. At this time, memory 3
Intra quantization table for addresses 0 to 63 (area A) of 2 and Non for addresses 64 to 127 (area B)
At the same time as writing the Intra quantization table, the same value is also written to the memory 43.

Hereinafter, the operation of the memory 31 will be mainly described in the case of the quantization processing. In MPEG processing,
The selector 42 selects a signal for switching between Intra and Non-Intra and becomes a switching signal for the selector 38.
At the time of Intra encoding, the selector 38 inputs the address signal generated by the address generator 39 to the memory 31 and reads the quantization table from the area A. At this time, a quantization table corresponding to the coordinate position (i, j) of the DCT coefficient 33 is input to the quantizer 5. At the time of Non-Intra encoding, the selector 38 inputs the address signal generated by the address generator 40 to the memory 31 and reads the quantization table from the area B. At this time, the quantizer 5 has D
A quantization table corresponding to the coordinate position (i, j) of the CT coefficient 33 is input. Note that the above (i, j) corresponds to FIG.
8 (b).

In the JPEG processing, the quantization table for the luminance signal is written in the area A of the memory 31 and the quantization table for the color difference signal is written in the area B of the memory 31 by the input 50 from the external CPU. At the time of quantization, a selector 42 selects a luminance signal / color difference signal switching signal, which becomes a selector 38 switching signal. At the time of encoding the luminance signal, the selector 38 inputs the address signal generated by the address generator 39 to the memory 31 and reads the quantization table from the area A. At this time, the quantizer 5 stores the coordinate position (i,
A quantization table corresponding to j) is input. At the time of encoding the color difference signal, the selector 38 inputs the address signal generated by the address generator 40 to the memory 31 and reads the quantization table from the area B. At this time, a quantization table corresponding to the coordinate position (i, j) of the DCT coefficient 33 is input to the quantizer 5. Note that the above (i, j) corresponds to FIG.
8 (b) and FIG. 8 (b).

Next, the operation of the inverse quantization process will be described focusing on the operation of the memory 43. In the MPEG decoding, the selector 49 selects a signal for switching between Intra and Non-Intra, and becomes a switching signal for the selector 45. At the time of Intra decoding, the selector 45 inputs the address signal generated by the address generator 46 to the memory 43, and reads the quantization table from the area A. At this time, a quantization table corresponding to the coordinate position (i, j) of the variable length decoding result 36 is input to the inverse quantizer 6. No
At the time of n-Intra decoding, the selector 45 inputs the address signal generated by the address generator 47 to the memory 43, and reads the quantization table from the area B. At this time, a quantization table corresponding to the coordinate position (i, j) of the variable length decoding result 36 is input to the inverse quantizer 6. Note that the above (i, j) corresponds to those shown in FIGS. 7B and 8B.

In the JPEG processing, the quantization table for the luminance signal is written in the area A of the memory 43, and the quantization table for the color difference signal is written in the area B of the memory 43 by the input 50 from the external CPU. At the time of inverse quantization, a selector 49 selects a switching signal of a luminance signal and a color difference signal, which is a switching signal of the selector 45. When decoding the luminance signal, the address signal generated by the address generator 46 is stored in the selector 43 by the selector 45.
, And the quantization table is read from the area A.
At this time, a quantization table corresponding to the coordinate position (i, j) of the variable length decoding result 36 is input to the inverse quantizer 6. When the color difference signal is decoded, the address generator 4 is selected by the selector 45.
7 is input to the memory 43, and the quantization table is read from the area B. At this time, the coordinate position (i, j) of the variable length decoding result 36 is stored in the inverse quantizer 6.
Is input. Note that (i,
j) corresponds to those shown in FIGS. 7B and 8B.

At the time of reset, MPEG processing, JP
In each of the EG processes, a corresponding default quantization table may be automatically loaded from the internal ROM into the memories 31 and 43 according to the mode.

In the sixth embodiment, the quantization table input from the external CPU is written. However, if the quantization table is written from the outside, a load is imposed on the external CPU. Since the quantization table used in the MPEG processing is a fixed value, the default quantization table and the inverse quantization table used in the MPEG processing are stored in the internal R of the encoding means 10.
OM, and from the internal ROM to the memories 31, 4
3, the load on the CPU can be prevented.

This loading is automatically performed from the internal ROM to the memory 31 according to the processing mode, such as at the time of resetting or switching between the respective modes of MPEG processing and JPEG processing. With this configuration, usability can be improved and it is convenient.

According to the sixth embodiment, by separately providing the quantizer / inverse quantizer, the quantization process and the inverse quantization process can be performed simultaneously and independently. In the MPEG encoding process that performs both processes, it is not necessary to perform both processes alternately at different times, and time constraints can be reduced.

By integrating FIGS. 1 and 2 above, a moving picture / still picture coding / decoding apparatus can be realized in the same manner. FIG.
And FIG. 4 can be integrated to realize a still image encoding / decoding device in the same manner.

Next, FIG. 12 shows an example of an image pickup apparatus using the coding apparatus or the decoding apparatus according to the first to sixth embodiments. 60 is a lens, 61 is an image sensor, 62 is an amplifier, 6
Reference numeral 3 denotes an analog-to-digital converter for converting an analog signal into a digital signal; 64, a signal processing circuit for processing a digital signal to generate an image signal; 65, a buffer memory for temporarily storing a video signal; CODEC which is an encoding device or a decoding device of the sixth embodiment,
67 is a memory such as a flash memory for storing a compressed image signal, 68 is a signal processing circuit 64, a buffer memory 6
5, a control circuit for controlling the CODEC 66, the memory 67, and the like; 69, an instruction unit for instructing power-on and switching between a moving image / still image mode; and 70, a common communication unit. The subject image is formed on the image sensor 61 through the lens, and is photoelectrically converted by the image sensor 61 to generate an electric signal. The electric signal is amplified by the amplifier 62, and the amplified analog electric signal is converted into a digital signal by the A / D converter 63. The digital signal is subjected to predetermined processing in a signal processing circuit 64 and is once written in a buffer memory 65. The image signal held in the buffer memory 65 is
The encoded data is generated by the DEC 66 to generate a compressed signal.
7 is stored. In this embodiment, the MPEG processing for performing the moving image processing and the JPE processing for performing the still image processing in the CODEC 66
Since some common circuits can be used in the G processing, the circuit scale itself can be reduced, but the imaging device itself can be reduced in weight, size, and power consumption.

Instead of writing the quantization table from the CPU 68 into the memory used for quantization shared by the moving image processing and the still image processing, a ROM storing the quantization table is provided in the CODEC 66. Is written in the memory in the CODEC 66, the load on the CPU 68 is reduced, and the inconvenience of the CPU processing the control of the imaging device becoming heavy or inoperable can be solved. At this time, the operation of the power switch and the operation of switching the still image / video mode are input from the instruction means 69 to the CPU 68, so that the quantization table is stored in the CODEC 66 when the power switch is turned on or when the still image / video mode is switched. It is also possible to adopt a configuration in which control is performed so that data is written into the memories 31 and 32 in the memory.

Further, at the time of still image processing, not only hardware processing but also variable coding processing of a variable table by software is performed, so that the compression ratio can be improved. Since the compression ratio can be improved, the number of still images that can be stored with the same storage capacity can be increased by using the encoding or decoding means described in the first to fourth embodiments in an imaging device. It is possible to improve the image quality of the compressed image even if the number of images is the same, which is advantageous for the imaging apparatus.

[0066]

According to the present invention, the circuit scale of a processing circuit for encoding and decoding moving images and still images can be reduced. Also, by using software processing together,
Encoding with a high degree of freedom according to an image can be realized.

[Brief description of the drawings]

FIG. 1 is a block diagram of an encoding apparatus that performs moving image encoding using an MPEG system and still image encoding using a JPEG system by sharing circuits as a first embodiment of the present invention.

FIG. 2 is a block diagram of a decoding apparatus that performs moving image decoding using the MPEG system and still image decoding using the JPEG system by sharing circuits as a second embodiment of the present invention.

FIG. 3 is a block diagram of a still image encoding device using a JPEG system according to a third embodiment of the present invention.

FIG. 4 is a block diagram of a still image decoding apparatus using a JPEG system according to a fourth embodiment of the present invention.

FIG. 5 is an explanatory diagram regarding image formats of [4: 2: 0] and [4: 2: 2].

FIG. 6 is an explanatory diagram illustrating a processing period in the image formats of [4: 2: 0] and [4: 2: 2].

FIG. 7 is an explanatory diagram of a quantization circuit in encoding moving images and still images.

FIG. 8 is an explanatory diagram of an inverse quantization circuit in decoding a moving image and a still image.

FIG. 9 is an explanatory diagram of a default quantization table in the MPEG standard.

FIG. 10 shows a fifth embodiment of the present invention, MPEG / J.
FIG. 3 is a block diagram of a quantization / dequantizer shared by PEG.

FIG. 11 shows a sixth embodiment of the present invention, MPEG / J.
FIG. 3 is a block diagram of a quantization / dequantizer shared by PEG.

FIG. 12 is a block diagram of an imaging apparatus according to a seventh embodiment of the present invention.

[Explanation of symbols]

 REFERENCE SIGNS LIST 1 block generator 2 subtractor 3, 14, 15 selector 4 DCT calculator 5 quantizer 6 inverse quantizer 7 IDCT calculator 8 adder 9 prediction memory 10 JPEG variable-length encoder (fixed table) 11 MPEG Variable length encoder 12 Hardware encoder 13 JPEG variable length encoder (variable table) 16 Stream multiplexer 17 Stream buffer

 ──────────────────────────────────────────────────続 き Continuing from the front page (72) Inventor Yukitoshi Tsuboi 5-20-1, Josuihonmachi, Kodaira-shi, Tokyo Inside the System LSI Development Center, Hitachi, Ltd. F-term (Reference) 5C022 AA13 AC69 5C053 FA08 GA11 GB22 GB26 GB29 GB34 GB36 GB38 KA03 KA05 KA08 KA22 KA25 5C059 KK07 KK50 MA00 MA23 MC11 MC38 ME01 PP01 PP04 TA14 RC14 TA58 TB07 TC18 TC24 TD06 TD11 TD15 UA02 UA05 UA25 UA32 UA33 5C078 AA09 BA22 BA32 BA57 CA01 CA26 CA31 DA01 DA02

Claims (25)

[Claims] 1. A storage device comprising: a rewritable storage unit; and a quantization unit that reads out a quantization table stored in the storage unit and performs quantization, and stores a value of the quantization table written in the storage unit into a moving image. An image coding apparatus characterized by performing quantization corresponding to both moving image processing and still image processing by changing between processing and still image processing. 2. An inverse quantization table, comprising: a rewritable storage means; and an inverse quantization means for reading an inverse quantization table stored in the storage means and performing inverse quantization. An image decoding apparatus characterized in that the value of is changed between moving image processing and still image processing to perform inverse quantization corresponding to both moving image processing and still image processing. 3. A rewritable storage unit, a quantization unit that reads a quantization table stored in the storage unit and performs quantization, and reads an inverse quantization table stored in the storage unit and performs inverse quantization. Inverse quantization means for performing both the moving image processing and the still image processing by changing the values of the quantization table and the inverse quantization table written in the storage means between the moving image processing and the still image processing. An image encoding / decoding apparatus characterized by performing quantization and inverse quantization corresponding to (1). 4. An image coding apparatus comprising: means for dividing an image into blocks; means for converting a block-divided image into a frequency domain; and quantization means for quantizing a signal converted into the frequency domain. In the above, the quantization means has a rewritable storage means, reads out the quantization table stored in the storage means, performs quantization, and stores the value of the quantization table written in the storage means in moving image processing and static An image coding apparatus characterized by performing quantization corresponding to both moving image processing and still image processing by changing between image processing and image processing. 5. A stream separating means for separating a compressed stream, a decoding means for decoding the separated stream, and an inverse quantization means for inversely quantizing data output from the decoding means. In the image decoding apparatus having the above, the quantization means has a rewritable storage means, reads out a quantization table stored in the storage means, performs inverse quantization, and writes an inverse quantum stored in the storage means. An image decoding apparatus characterized in that inverse quantization corresponding to both moving image processing and still image processing is performed by changing the value of a conversion table between moving image processing and still image processing. 6. A means for dividing an image into blocks, a means for converting a blocked image into a frequency domain, a quantization means for quantizing a signal converted into a frequency domain, and a stream separation for separating a compressed stream. Means; a decoding means for decoding the separated stream; and an inverse quantization means for inversely quantizing the data output from the decoding means. The quantization means or the inverse quantization means has a rewritable storage means, reads out the quantization table or the inverse quantization table stored in the storage means, performs quantization or inverse quantization, and writes the data in the storage means. Quantization or inverse quantization corresponding to both moving image processing and still image processing is performed by changing the value of the quantization table or inverse quantization table used for moving image processing and still image processing. An image encoding / decoding apparatus characterized by the above-mentioned. 7. An inverse quantization unit for inversely quantizing data output from the quantization unit, independent of the quantization unit, wherein a quantization process by the quantization unit and an inverse quantization by the inverse quantization unit are performed. The image coding apparatus according to claim 1, wherein the inverse quantization processing can be performed simultaneously. 8. A storage device according to claim 1, further comprising a ROM for storing the quantized data to be written in said storage means, wherein said quantized data is written from said ROM to said storage means. An image encoding device according to any one of claims 1 to 3. 9. The image encoding apparatus according to claim 8, wherein the quantization data is written from the ROM in accordance with an external instruction. 10. An image coding apparatus having a quantization unit and an inverse quantization unit independently, and capable of simultaneously executing a quantization process and an inverse quantization process. 11. A first storage means for storing a quantization table, a second storage means for storing an inverse quantization table, and reading out a quantization table stored in the first storage means. A first storage unit and a second storage unit, comprising: a quantization unit that performs quantization; and an inverse quantization unit that reads an inverse quantization table stored in the second storage unit and performs inverse quantization. The image encoding apparatus is independent of the means, and is capable of simultaneously executing the quantization process and the inverse quantization process. 12. An image coding apparatus comprising: means for dividing an image into blocks; means for converting a block-divided image into a frequency domain; and means for quantizing a signal converted to a frequency domain. A first variable-length coding unit that performs variable-length coding on a quantized signal using a fixed first coding table, and a second coding table that can be arbitrarily set. An image code comprising: a second variable length coding unit that performs variable length coding; and a selection unit that selects between the first variable length coding unit and the second variable length coding unit. Device. 13. The method according to claim 1, wherein the selection means selects the first variable-length coding means and the second variable-length coding means according to an external setting or image characteristics.
3. The image encoding device according to 2. 14. The image code according to claim 12, wherein said first selector selects one of the first variable-length encoder and the second variable-length encoder having a higher compression rate. Device. 15. The image encoding apparatus according to claim 12, wherein said second encoding means performs the processing by software processing. 16. The apparatus according to claim 16, further comprising: means for outputting from said first variable length coding means and means for outputting from said quantization means, wherein said quantization means is always outputted or outputted in accordance with a mode. An image encoding device, comprising: means for selecting 17. A fixed stream decoding means for separating a compressed stream, a decoding means for decoding the separated stream, and an inverse quantization means for inversely quantizing data output from the decoding means. Means for performing a first variable length decoding according to the coding table obtained, and means for performing a second variable length decoding using an arbitrary coding table included in a signal to be decoded, An image decoding apparatus comprising: means for performing first variable-length decoding; and selecting means for selecting means for performing second variable-length decoding. 18. The method as claimed in claim 18, wherein the selection means selects the first variable length decoding means and the second variable length decoding means according to the type of the signal to be decoded. 18. The image decoding device according to item 17. 19. The image decoding apparatus according to claim 17, wherein said first decoding means is performed by hardware processing, and said second decoding means is performed by software processing. Device. 20. An input means to which said stream is input, and an input means to which an output of said second variable-length decoding means are input, and means for performing first variable-length decoding.
17. The image decoding apparatus according to claim 16, wherein said input means is selected in accordance with a mode selection between a first mode and a second mode for performing a second variable length decoding. 21. An imaging apparatus comprising: a lens; imaging means for photoelectrically converting an optical image formed by the lens; and compression means for compressing an image signal output from the imaging means. Rewritable storage means, and a quantization means for reading out a quantization table stored in the storage means and performing quantization, and performs a moving image processing and a still image processing on the values of the quantization table written in the storage means. An image pickup apparatus characterized in that the quantization corresponding to both the moving image processing and the still image processing is performed by a single compression unit by changing the above. 22. The apparatus according to claim 21, further comprising power-on means for instructing power-on, wherein the quantization table is written in the storage means in response to an operation of the power-on means. Imaging device. 23. The apparatus according to claim 21, wherein the compression unit has a ROM in which the quantization table is stored in advance, and writes the quantization table stored in the ROM into the storage unit. Imaging device. 24. An imaging apparatus comprising: a lens; imaging means for photoelectrically converting an optical image formed by the lens; and compression means for compressing an image signal output from the imaging means. A first storage unit in which a quantization table is stored, a second storage unit in which an inverse quantization table is stored, and a quantization table read out from the first storage unit to perform quantization. A quantizing unit, and an inverse quantizing unit that reads the inverse quantization table stored in the second storage unit and performs inverse quantization, wherein the first storage unit and the second storage unit are independent. And an imaging apparatus capable of simultaneously executing the quantization process and the inverse quantization process. 25. An image pickup apparatus comprising: a lens; image pickup means for photoelectrically converting an optical image formed by the lens; and compression means for compressing an image signal output from the image pickup means. Means for performing variable-length coding on a quantized signal using a fixed first coding table, and means for performing variable-length coding using an arbitrarily configurable second coding table An imaging apparatus comprising: a first variable-length encoding unit and a second variable-length encoding unit that are selected by a characteristic of an image or an instruction of a user.