JPH04207266A - Encoder and encoding method for image data - Google Patents

Abstract

PURPOSE:To perform the encoding of a wide spatial frequency image so as to be kept within constant encode quantity for constant processing time with simple constitution by predicting the optimum quantization width by calculated encode quantity and quantization width, etc., based on variable length encode output when statistical processing is performed, and controlling quantization classified by every chrominance component when encode processing is performed. CONSTITUTION:A quantization width prediction circuit predicts the optimum quantization width classified by every chrominance component at an encoder circuit 80 which encodes the image pickup image signal of an electronic camera by the calculated encode quantity classified by every chrominance component and the quantization width by an encode quantity calculation circuit 14 based on the output of a variable length encoder circuit 8 and a total encode quantity target value from a control circuit 18 replying to target encode quantity setting information from a control circuit 90. The quantization and encoding classified by every chrominance component are performed based on predicted width, and the encode quantity can be controlled for the wide spatial frequency without impairing picture quality with simple constitution, and encoding processing can be performed so as to be kept within the constant encode quantity for constant processing time.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to an encoding device and encoding method for highly compression encoding image data to a predetermined capacity.

[Conventional technology]

When an image signal captured by a solid-state imaging device such as a CCD is stored as digital data in a storage device such as a memory card, magnetic disk, or magnetic tape, the amount of data is enormous, so many frames are required. In order to record images within a limited storage capacity, it is necessary to perform some kind of highly efficient compression on the obtained image signal data. Furthermore, in digital electronic still cameras, captured images are stored as digital data on storage media such as memory cards, magnetic disks, or magnetic tapes instead of silver halide films, so they can be stored as digital data on storage media such as memory cards, magnetic disks, or magnetic tapes. The number of images that can be recorded on one roll of magnetic tape device is specified, and the recording of this specified number of images must be guaranteed, and the time required for data recording and reproduction processing is short and constant. Is it necessary?

Similarly, digital VTR (video tape recorder),
When recording moving images such as digital moving image files, it is necessary to be able to record various amounts of frames without being affected by the amount of image data per frame.

In other words, whether it is a still image or a moving image, it is necessary to be able to reliably record the required number of frames, and the time required for data recording and reproduction processing must be short and constant. .

A coding method that combines orthogonal transform coding and variable length coding is widely known as a highly efficient image data compression method.

A typical example is the method being considered in the international standardization of still image coding.

An outline of this method will be explained below. First, image data is divided into blocks of a predetermined size, and a two-dimensional DCT (discrete cosine transform) is performed as orthogonal transformation for each divided block. Next, linear quantization is performed according to each frequency component, and this quantized value is encoded as a variable length code.
Perform human encoding. At this time, regarding the DC component, the difference value between the DC component and the DC component of the neighboring program is encoded by /X human coding. The AC component is scanned from a low frequency component to a high frequency component called a zig-sag scan, and two-dimensional /% Human encoding is performed from the consecutive number of invalid (value or O) components and the value of the subsequent valid component. do.

Since Huffman encoding, which is variable length encoding, is used, the amount of code for this basic part is not constant for each image.

Therefore, the following method has been proposed as a method for controlling the amount of code. First, at the same time as processing the basic part, the total amount of codes generated for the entire screen is calculated.

From this total code amount and the target code amount, the optimum quantization width for approaching the target code amount for the DCT coefficient is predicted. Next, using this quantization width, the processing after quantization of the basic portion is repeated.

Then, from the total code amount generated this time, the total code amount generated last time, and the target code amount, the optimum quantization width to bring the target code amount closer again is predicted.

If the predicted quantization width and the previous quantization width are sufficiently close and the total amount of code generated this time is smaller than the target code amount, the process is terminated and the code is output. If not, the process is repeated using a new quantization width.

The quantization width is obtained by preparing a quantization matrix, which is a basic form representing relative quantization characteristics for each frequency component, and multiplying this quantization matrix by a quantization coefficient to obtain the necessary quantization width. Specifically, first, the above-mentioned basic part is quantized using the quantization width obtained using standard quantization coefficients, and then this is variable-length encoded, and the information on the total code amount obtained by this is , a comparison is made with the preset target total code amount, which is the limit that should be accommodated, and when the target total code amount is reached, the final encoding process is performed using that quantization width, and the target total code amount is reached. If the amount does not fall within the amount, the optimal quantization coefficient that approaches the target total code amount is determined by linear prediction from the generated total code amount and the target total code amount, and
A more optimized quantization width is calculated from the obtained quantization coefficient and quantization matrix, and this is used to perform the final encoding process. The quantization width is changed using this method.

However, with the above-mentioned code amount control method, the number of times the bus, which is the basic part of encoding, is repeated differs depending on the image, so not only does the processing time become unstable, but also - Long processing time is required for boat fishing. There were drawbacks.

Therefore, the present inventors proposed the following method as a method to solve this problem. (Refer to Japanese Patent Application No. 1-283761 and Japanese Patent Application No. 2-137222.) This proposed method is a compression method that combines orthogonal transform and variable-length coding, and is designed to control the amount of generated code. Divide the sampled image signal stored in the memory into blocks, perform orthogonal transformation for each divided block, quantize this transform output with a provisional quantization width, and then convert the quantized output to is variable-length encoded, and calculates the generated code amount for each block and the total generated code amount for the entire image, and then calculates the provisional quantization width, the total generated code amount, and the target total code amount. From this, predict the new quantization width. Further, the allocated code amount for each block is calculated from the generated code amount for each block, the total generated code amount, and the target total code amount. Then, the image signal in the image memory is divided into blocks, orthogonally transformed, quantized, and variable length encoded again using the new quantization width, and if the generated code amount of each block exceeds the allocated code amount of each block, stops variable-length encoding midway through and moves on to processing the next block. As a result, if the total generated code amount of the entire image does not exceed the target total code amount, one frame of image data (one frame of The image is 720 x 57B pixels) is divided into blocks of a predetermined size (for example, blocks A, B, C, etc. each consisting of 8 x 8 pixels), and as shown in (b), the divided A two-dimensional DCT (discrete cosine transform) is performed as an orthogonal transform for each block, and the data is sequentially stored on an 8×8 matrix. When image data is viewed on a two-dimensional plane, it has a spatial frequency that is frequency information based on the distribution of gradation information.

Therefore, by performing the above DCT, the image data is converted into a direct current component DC and an alternating current component AC, and data indicating the value of the direct current component DC is placed at the origin position (0,0 position) on the 8×8 matrix. At the 0.7 position, there is data indicating the maximum frequency value of the alternating current component AC in the horizontal axis direction, and
Data indicating the maximum frequency value of the alternating current component AC in the vertical axis direction is stored at positions 7 and 0, and data indicating the maximum frequency value of the alternating current component AC in the diagonal direction is stored at positions 7 and 7. Then, the frequency data in the directions related by the respective coordinate positions are stored in a form in which higher frequencies appear sequentially from the origin side.

Next, by dividing the stored data at each coordinate position in this matrix by the quantization width for each frequency component,
Linear quantization is performed according to each frequency component (C), and the quantized values are subjected to Noffman encoding as variable length encoding. At this time, regarding the direct current component DC, the nearby professional
The difference value from the DC component of the tsuku is expressed by a group number (number of additional bits) and additional bits, the group number is Nofman encoded, and the obtained code word and additional bits are combined to form encoded data ( di, d2. el, e2
).

Coefficients that are also valid for the alternating current component AC (values other than 0°) are expressed by group numbers and additional bits.

Therefore, AC component formula 〇 performs a scan called zigzag scan from low frequency components to high frequency components, and calculates the number of consecutive invalid (value or "0") components (number of runs of zero) and the following valid components. Two-dimensional Huffman encoding is performed using the group number of the component value, and the resulting code word and additional bits are combined to form encoded data.

Huffman encoding is the above DC component D per frame image.
Centering on the peak frequency of occurrence in the data distribution of C and alternating current component AC, bits are allocated in such a way that the center has fewer data bits, and the closer it is to the periphery, the larger the number of bits. This is done by encoding data to obtain a code word.

The above is the basic part of this method.

If only this basic part is used, the code will not be constant for each image because Huffman coding, which is variable length coding, is used, so the code amount is controlled, for example, as follows.

First, using a provisional quantization coefficient, the basic part is processed with the quantization width for each frequency component obtained by multiplying the predetermined quantization matrix and the quantization coefficient, and at the same time The total amount of generated codes (total number of bits) is determined (g). The optimum quantization coefficient for approaching the target code amount for the DCT coefficient is predicted from this total code amount, the target code amount, and the temporary quantization coefficient used (h). Next, using this quantization coefficient (j), the above-described process after quantization of the basic part is repeated. Then, from the total code amount generated this time, the total code amount generated last time, the target code amount, the quantization coefficient used this time, and the quantization coefficient used last time, it is possible to approach the target code amount again. The optimal quantization coefficients are determined using Newton-Raphson iteration.
Predict by n).

If the predicted quantization coefficient and the previous quantization coefficient are sufficiently close, and the total amount of code generated this time is less than the target amount of code, the process is terminated, and the code amount generated this time is Output the data and store it on the memory card (f
). Otherwise, the quantization coefficient is changed and the process is repeated using this new quantization coefficient.

However, with the method of controlling the amount of code described above, the number of times the basic encoding pass is repeated differs depending on the image, which not only makes the processing time unstable, but also generally requires a long processing time. There was. Therefore, the present inventors solved this problem using the following method.

First, using a provisional quantization coefficient, the basic part is processed with the quantization width for each frequency component obtained by multiplying the determined quantization matrix and the quantization coefficient, and at the same time, the entire screen is processed. The total amount of generated codes (total number of bits) and the amount of generated codes for each block are determined (gl). The optimum quantization coefficient for approaching the target code amount for the DCT coefficient is predicted from this total code amount, the target code amount, and the temporary quantization coefficient used (hl). Further, the allocated code amount for each block is determined by proportional allocation from the total code amount, the target code amount, and the generated code amount for each block (h2). Next, using the predicted quantization coefficients, (1)
, repeat the basic processing described above. In other words, the image signal in the image memory is divided into blocks, orthogonally transformed, quantized,
When variable length encoding is performed and the amount of code generated for each block exceeds the amount of allocated code for each block, variable length encoding is stopped midway and processing proceeds to the next block.

After completing the above processing for all blocks, the encoded data generated this time is output and stored in a memory card (0゜ [Problem to be solved by the invention]) As mentioned above, for example, a digital electronic still In cameras, etc., the number of images that can be recorded on one memory card, magnetic disk device, or magnetic tape must be guaranteed, so image data is compressed and recorded, but it is difficult to operate. Above all, the processing time needs to be as short as possible and constant.

It is also desired that image data can be compressed with high efficiency. These are items that are required not only for digital electronic still cameras but also for other applications.

As a compression method that satisfies these requirements, there is the above-mentioned International Standards Laboratory method. In this method, the method that combines block-by-block orthogonal transformation and variable-length coding as exemplified in the basic section above cannot process image data. However, since variable-length encoding is used, the number of images that can be stored on a recording medium such as a memory cart, magnetic disk, or magnetic tape remains constant. The problem was that it was not fixed.

In addition, as a conventional technique to solve this problem, there is a code amount control method that is being considered in the international standardization of still image coding mentioned above, but in this method, the number of passes of the basic part of encoding is repeated. Not only is the processing time undefined because it varies depending on the image, but it also has the drawback of generally requiring a long processing time.

As a method to solve this problem, the present inventors proposed Japanese Patent Application No. 1-28376 and Japanese Patent Application No. 2-137222.
There is a number system. The second proposed method usually gives very good results, but depending on the system to which it is applied, the image quality may deteriorate slightly for images with particularly low or particularly high spatial frequencies, and the amount of code generated may be In very rare cases, the difference in the amount of code becomes slightly larger than in the normal case.

The cause of this will be explained. As mentioned above, when expressing the quantization width by multiplying the quantization matrix, which is the basic form expressing the relative quantization characteristics for each frequency component, by the quantization coefficient, there is a difference between the quantization coefficient and the amount of generated code. has approximately the following relationship.

log BR-a log sp+ b
=・(1) Here, BR is the amount of code per pixel (bit rate), SF is the quantization coefficient, a and b are constants, a is particularly dependent on the system, and b is particularly large on the image. Dependent.

When quantizing a color image separated into color components, a plurality of quantization matrices, for example, two quantization matrices for luminance and color difference, are used for the color components, and
If we use a common quantization coefficient, log BR= a log SP+ b
-(2) y y
ylog BR-a log SP+ b
-(3) The relationship is like CCC. In addition, in each of the above formulas, BR.

y and c are added to a and b to distinguish between brightness and color difference. Here, if a = ao, then = a
log SP+ b - (4), and it can be seen that the allocated code amount for each block can be determined by proportionally allocating the target code amount according to the statistics.

In the normal case, aha, but the system y
Depending on c, in very rare cases the size of a and a may be quite large y
c It may be different.

In other words, the total code amount log BR is plotted on the Y axis, and the quantization coefficient l
og SF on the X axis and the relationship between equations (2) and (3) is shown in Figure 10, and the quantization coefficient log SF
The total code amount according to log BR is slightly different between the luminance system and the color difference system. Looking at Figure 10, when the quantization coefficient during encoding, that is, the quantization coefficient predicted to be the optimal value, is particularly large or small,
In other words, especially in the case of images with high or low spatial frequencies, if the allocated code amount is determined by proportionally allocating the target code amount according to the statistics of each block, the predicted quantization When coding is actually performed using the coefficients, the allocated code amount becomes insufficient for the luminance component block, and becomes redundant for the chrominance component block, or vice versa.

Whether the allocated code amount for the luminance component is insufficient or the allocated code amount for the chrominance component is insufficient depends on the magnitude relationship between a and a and the magnitude of the predicted quantization coefficient y c .

Due to the above reasons, depending on the applied system, the image quality may deteriorate slightly for images with particularly low or particularly high spatial frequencies, and the gap between the generated code amount and the target code amount may be slightly larger than in normal cases. was there.

Therefore, it is an object of the present invention to encode an image so that it stays within a certain amount of code within a certain processing time without degrading the image quality, even if the image has a biased spatial frequency distribution. An object of the present invention is to provide an image data encoding device and encoding method that can perform image data encoding.

[Means to solve the problem]

In order to achieve the above object, the present invention is configured as follows.

That is, each color image data separated by color component is divided into blocks of a predetermined number of pixels, and each divided block by color component is sequentially subjected to orthogonal transformation to be converted to data by frequency component. The quantized data for each frequency component is quantized by the quantization means, and then this quantized output is given to the variable length encoding means for variable length encoding. Quantize with a predetermined quantization width and perform variable length coding, perform statistical processing to obtain the code amount for each block, and then calculate the code amount for each block obtained by this statistical processing and the total code amount target per image. The optimum quantization width and the allocated code amount for each block are determined based on the total code amount target value, and the transform output is quantized for each block according to the determined optimum quantization width. In an encoding apparatus that performs a coding process in which the quantized output is sequentially variable-length encoded while controlling the code amount for each block according to the allocated code amount for each block, the first step is to perform the statistical processing. calculation means for calculating the code amount of each block for each color component based on the variable length encoded output at the time; quantization width prediction means for predicting an optimum quantization width for each color component based on the quantization width; and using the optimum quantization width for each color component predicted by the quantization width prediction means during the encoding process. and a control means for controlling to perform quantization for each color component.

Second, a calculation means for calculating the code amount of each block for each color component based on the variable length encoded output during the statistical processing, and the total code amount target value and the calculated value of the calculation means and the quantization width prediction means for predicting an optimal quantization width for each color component based on the provisional quantization width used for statistical processing;
determining means for allocating the total code amount target value according to the calculated value of the calculating means and determining a reference code amount for each color component side number block; and for each color component determined by the determining means during encoding processing. Using the standard code amount of each block, the code amount generated by encoding in the processed block processed before the next block to be encoded among the blocks and the allocated code amount in the processed block. and adding this to the reference code amount of the block to be processed, and assigning means for allocating the obtained code amount as the allocated code amount in the block to be processed, and encoding in the block to be processed. In order not to exceed the allocated code amount allocated by the allocation means in the processing,
an abort means for aborting the variable length encoding; and a control means for controlling to perform quantization for each color component using the optimum quantization width for each color component predicted by the quantization width prediction means during the encoding process. It is composed of:

Thirdly, for each of the above first and second configurations,
Furthermore, the apparatus is configured to include compression rate selection means for selecting a target compression rate and determining the target code amount to be accommodated and the provisional quantization width according to the selected compression rate.

Fourthly, in each of the first and third configurations, the quantization means may perform quantization at the time of the encoding process to perform quantization on each of the frequency component data obtained by the orthogonal transformation. In addition, the variable length encoding means uses the data of the previous processing block for the DC component among the frequency component data of the image data in units of blocks obtained for each frequency component by the orthogonal transformation. a DC component encoding processing unit that obtains a difference with the DC component at and variable-length encoding this difference; and after the encoding processing unit finishes processing the DC component, variable-length encoding the AC component; An encoding processing section for alternating current components is provided, and the encoding aborting means is provided only in the encoding processing section for alternating current components.

Fifth, the color image data separated by color component is divided into blocks each having a predetermined number of pixels, and each divided block is sequentially subjected to orthogonal transformation, and then the transformation output is quantized by quantization means. After that, this quantized output is given to a variable length encoding means to perform variable length encoding, and when encoding data, it is calculated from the total amount of code to be stored and the statistics for each block determined in advance. When performing variable length encoding while controlling the code amount for each block according to the determined allocated code amount for each block, the quantization is performed using a provisional quantization width, and the resulting quantization A first step of performing encoding processing on the output and checking the code amount of the entire image, the code amount for each color component, and the code amount for each block for each color component, and the code amount of the entire image obtained in this first step. The second step is to pre-determine the quantization width necessary for optimization based on the amount of code for each color component and the amount of code for each block for each color component. a third step of determining a standard allocated code amount for each block; a fourth step of performing the quantization for each block for each color component using the predicted quantization width for each color component; a fifth step of variable-length encoding the binary output of each block for each color component in step 4 within the allocated code amount for that block; and a fifth step of variable-length encoding the current output in the fifth step. Find the difference between the amount of code generated by variable-length encoding in the block to be processed immediately before and the amount of code allocated to this block, and use this as the carryover amount for the block to be processed currently. and a sixth step of adding the reference allocated code amount to the allocated code amount for the block to be currently processed.

Sixthly, in the fifth method, by setting the compression rate before the first step, the target total code amount to be accommodated and provisional quantization are determined according to the set compression rate. Configure by adding a step to determine the width.

Furthermore, seventhly, in the fifth and sixth methods, the provisional quantization width is a predetermined value according to the selected total code amount per screen.

[For production]

In such a configuration, each color image data separated by color component is divided into blocks of a predetermined number of pixels,
Orthogonal transformation is sequentially performed on each divided block for each color component to convert it into frequency-classified data, and this transformed frequency-classified data is quantized by a quantization means, and then this quantized output is Statistics to obtain the code amount for each block by applying it to a variable length encoding means and performing variable length encoding, and when encoding data, quantize it with a provisionally determined quantization width and perform variable length encoding. Then, from the code amount for each block obtained through this statistical processing and the total code amount target value per image, the optimal quantization width suitable for keeping within this total code amount target value and the optimal quantization width for each block are determined. The allocated code amount is determined, and the transform output is quantized for each block according to the determined optimal quantization width, and the code amount is controlled for each block based on the quantized output according to the allocated code amount for each block. Alternatively, in the case of the first configuration, the calculating means calculates the code amount of each block for each color component based on the variable length encoding output during the statistical processing. and the quantization width prediction means predicts the optimal quantization width for each color component based on the total code amount target value, the calculated value of the calculation means, and the provisional quantization width used in the statistical processing. . Then, the control means performs control to perform quantization for each color component using the optimum quantization width for each color component predicted by the quantization width prediction means during the encoding process.

The above describes the total generated code amount for the entire screen based on quantization using a provisional quantization width during statistical processing, the generated code amount for each color component, and the state of the generated code amount for each block for each color component. The optimal quantization width for each color component is predicted from this information and the target total code amount, and when encoding each block of each color component in the encoding process, the optimal quantization width corresponding to the generated code amount of the color component is predicted. Since the quantization width can be determined and quantization can be performed using the optimal quantization width, the amount of generated code is optimized, which contributes to improving image quality. The time required can be constant.

In the case of the second configuration, the calculating means calculates the code amount of each block for each color component based on the variable length encoded output during the statistical processing, and the determining means calculates the total code amount target value based on the above-mentioned total code amount target value. It is distributed according to the calculated value, and the reference code amount of each block for each color component is determined. Further, the quantization width prediction means predicts the optimum quantization width for each color component based on the total code amount target value, the calculated value of the calculation means, and the provisional quantization width used in the statistical processing, During the encoding process, the allocation means uses the reference code amount of each block for each color component determined by the determination means, and among the blocks, a processed block processed before the next target block to be encoded is used. Find the difference between the code amount generated by encoding and the allocated code amount for the processed block, add this to the allocated code amount of the processing target block, and assign it as the allocated code amount for the processing target block. . Then, the aborting means aborts the variable length encoding so as not to exceed the allocated code amount allocated by the allocating means during the encoding process on the processing target block.

On the other hand, the control means performs control to perform quantization for each color component using the optimum quantization width for each color component predicted by the quantization width prediction means during the encoding process.

In the above configuration of Section 2, the total generated code amount for the entire screen based on quantization using a provisional quantization width during statistical processing, the generated code amount for each color component, and the generated code for each block for each color component. Knowing the state of the amount, predicting the optimal quantization width for each color component from this information, allocating the target total code amount according to the amount of generated code for each color component, and using this. The standard allocated code amount for each block is determined according to the generated code amount of each block for each color component, and when encoding each block in the encoding process, the standard allocated code amount for that block is used for the previous block. The difference between the code amount in the encoding and the allocated code amount is added as a carryover amount to obtain the allocated code amount, and encoding is performed within the range of this allocated code amount.Based on the data during statistical processing. , the standard allocated code amount for each block for each color component is determined by rationally distributing it, and the allocated code amount for the previously processed block is carried over and used as the code for subsequent blocks. Even for color components that require a large amount of code, there are fewer interruptions in coding for each block, contributing to improved image quality and reducing the number of times the coding process is performed. By setting , the time required for encoding can be kept constant.

In addition, in the case of the third configuration, a compression ratio selection means is further provided in the first and second cases,
When a target compression rate is selected, the compression rate selection means determines the target code amount to be accommodated and the provisional quantization width according to the selected compression rate.

Therefore, in this method, in addition to the functions and effects of the first and second configurations, the target compression rate can be selected, and the provisional quantization width and the target per image to be accommodated can be selected accordingly. Since the amount of code is determined and statistical processing is performed, an advantage is obtained that the optimum quantization width that fits within the desired amount of code can be found with higher accuracy.

Further, in the case of the fourth configuration, in each of the first and third configurations, the function of the variable length encoding means is applied to the DC component of the image data in units of blocks obtained by the orthogonal transformation. Obtain the difference from the DC component in the previous processing block and variable-length encode this difference. After the encoding processing unit finishes processing the DC component, variable-length encode the AC component and perform the encoding. A configuration is used in which truncation is performed only in the encoding process for AC components. Therefore, the quantization means performs quantization during the encoding process on data of low frequency components among the frequency component data obtained by the orthogonal transformation, and the variable length encoding means Of the frequency component data of the image data in each block obtained by the orthogonal transformation for each frequency component, for the DC component, a difference from the DC component in the previous processing block is obtained, and this difference is variable-length coded; After the encoding process for the DC component is completed, the AC component is subjected to variable length encoding, and the aborting of the encoding is applied only to the AC component.

As a result, in addition to the first to third cases above, by terminating the encoding only for the alternating current component during encoding processing, it is possible to limit the terminating to the high frequency region, which has little visual impact. Deterioration of image quality can be prevented.

In the fifth case, the color image data separated by color component is divided into blocks each having a predetermined number of pixels, orthogonal transformation is sequentially performed for each divided block, and the output of this transformation is quantized by a quantization means, After that, this quantized output is given to a variable length encoding means for variable length encoding, and when encoding data, the quantization output is determined from the total amount of code to be stored and the statistics for each block determined in advance. When performing variable length encoding while controlling the code amount for each block according to the allocated code amount for each block, the quantization is performed using a provisional quantization width in the first step, and the resulting gain is The code amount of the entire image, the code amount of the entire image, and the code amount of each block for each color component are checked, and in the second step, the entire image obtained in the first step is encoded. The quantization width for each color component necessary for optimization is predicted from the code amount and the code amount for each color component, and in the third step, the standard allocated code amount for each block is calculated from the code amount for each block, In the fourth step, the quantization is performed for each block of each color component using the predicted quantization width of each color component. and,
In the fifth step, the quantized output for each block in the fourth step is variable-length encoded within the allocated code amount for that block. The allocated code amount is determined by calculating the difference between the code amount due to variable length encoding of the block processed immediately before the current block to be subjected to variable length encoding processing and the allocated code amount for this block, and carrying this forward. The amount added to the reference allocated code amount for the block to be currently processed is applied.

In this way, in the fifth case, the total generated code amount for the entire screen based on quantization using a provisional quantization width, the generated code amount for each color component, and the generated code amount for each block for each color component are calculated. Knowing the state of the code amount, predicting the optimum quantization width for each color component from this information and the target total code amount to be stored, and also calculating the target value of the total code amount from the generated code amount for each color component. The standard allocation code amount for each block for each color component is determined from the allocated code amount and the generated code amount for each block for each color component, and the code for each block in the encoding process is determined. When converting, the difference between the coding amount in the previous block and the allocated code amount is added to the standard allocated code amount of the block as a carryover seedling to obtain the allocated code amount, and the code is allocated within the allocated code amount. Based on the data during statistical processing, the optimal quantization width for each color component is determined rationally, and the standard allocated code amount for each block for each color component is rationally determined. In addition, the amount of allocated code for the previously processed block can be carried forward and used for encoding subsequent blocks, so a large amount of code is not required. Even if the color component is a color component, the number of coding interruptions in each block is reduced, which contributes to improved image quality, and if the number of coding processes is set to a prescribed number, the time required for coding can be kept constant.

In the sixth case, by setting the compression rate, the target total code amount and provisional quantization width to be accommodated are determined according to the set compression rate, and then, in the fifth case. Implement the method.

In this way, first, the status of the total generated code amount for the entire screen based on quantization using a provisional quantization width, the generated code amount for each color component, and the generated code amount for each block for each color component is determined. From this information and the target total code amount to be stored, predict the optimal quantization width for each color component, allocate the target total code amount according to the generated code amount for each color component, and further Based on the allocated code amount and the generated code amount of each block for each color component, the standard allocated code amount for each block for each color component is determined, and when encoding each block in the encoding process, the standard for that block is determined. The difference between the code amount in the encoding of the previous block and the allocated code amount is added to the allocated code amount as a carryover amount to obtain the allocated code amount, and encoding is performed within the allocated code amount. Based on the data at the time of processing using the provisional quantization width, the standard allocated code amount for each block for each color component is rationally allocated and determined, and the previously processed block Since the allocated code amount for the next block can be carried over and used for encoding subsequent blocks, even for color components that require a large amount of code, coding is less likely to be interrupted in each block. This contributes to improving image quality, and if the number of encoding processes is set to a prescribed number, the time required for encoding can be kept constant.

Furthermore, in the seventh case, in the fifth and sixth cases, the provisional quantization width uses a predetermined value according to the selected total code amount per screen.

As a result, the provisional quantization width becomes close to the optimum value corresponding to the total number of codes per screen to be accommodated, and more appropriate compression encoding becomes possible.

Obtain the statistics necessary to predict the appropriate quantization width and determine the standard allocated code amount for each block for each color component, and start the encoding process based on this information, sequentially checking the encoded output. ,
Variable so as to fit within the allocated code amount of the block to be processed, which is obtained by combining the standard allocated code amount of the block to be processed and the excess or deficiency of the allocated code amount in the processing block before that block. By discontinuing long encoding, the code amount is controlled for each block, and an encoded output within the desired code amount is obtained as the final output.

This is based on the following idea.

As explained in the conventional example, when the quantization coefficient during encoding is particularly large or small, in other words, in the case of an image with a particularly high or low spatial frequency, the target code amount is determined according to the statistics of each block. When the allocated code amount is determined by proportional allocation of Yes, or vice versa.

Then, the encoding of the color component for which the amount of code is insufficient is aborted midway, and the image quality naturally deteriorates. Whether or not such a phenomenon occurs, whether or not the allocated code amount for the luminance component in the color component is insufficient, and whether or not the allocated code amount for the chrominance component in the color component is insufficient can be determined using the formula (2 ) of a
This can be easily determined based on the magnitude relationship of a in equation (3) and the magnitude of the predicted quantization coefficient.
In other words, it depends on the system and the image. Therefore, the excess code amount allocated for the block that has already been encoded is used as the allocated code amount for subsequent blocks.
From the above-mentioned relationship, after predicting the quantization width for each color component so that the ratio of the code amount for each color component does not change, encoding is performed while controlling the code amount for each block according to the prediction.

By using this predicted quantization width for each color component, the ratio of the generated code amount for each color component is the same as the ratio when the statistics were obtained, so determining the allocated code amount for each block is simply done using statistics. It is only necessary to proportionately distribute the target code amount according to the amount, and good results can be obtained for each color component.

The quantization width for each color component can be predicted by linear prediction using equations (2) and (3).

Therefore, according to the present invention, it becomes possible to encode images with a wide range of spatial frequencies within a certain amount of code within a certain processing time using a simple circuit without degrading the image quality as much as possible.

〔Example〕

Embodiments of the present invention will be described below with reference to the drawings. When compressing and encoding color image data,
When quantizing with the predicted optimal quantization coefficient and encoding it, if each color component such as luminance component or chrominance component is encoded so that it falls within the specified code amount range, depending on the color component, the code may be different. There may be cases where there is a surplus in the amount, but there is a shortage on the other side, and in such cases, the situation where the high frequency components are cut on the side of the insufficient color component occurs one after another, and on the other side, the high frequency components are cut. An unreasonable situation arises in which the amount of code becomes too large rather than causing the situation to occur. The present invention determines this by proportionally distributing the allocated code amount for each color component from the total generated code amount and the generated code amount for each color component, and determines this by proportionally distributing the allocated code amount for each color component. Appropriately allocate so that no components occur,
This makes it possible to rationally allocate the allocated code amount, thereby improving image quality.

First, in order to make the present invention easier to understand, the basic idea of the present invention will be explained. That is, the present invention, for example,
One color image (screen) is divided into color components, the color image data for each color component is divided into blocks of a predetermined number of pixels, and each color component block is sequentially divided into data of the constituent pixels of that block in block units. is orthogonally transformed to separate it into frequency components, which is further quantized, compression encoded, and encoded so that the total code amount of each compression encoded color component falls within the target code amount. Statistical processing is performed on the separated color image data, and this statistical processing predicts the optimal quantization width for each color component.
Statistics necessary for determining the proportional allocation of the allocated code amount for each color component based on the generated code amount for each color component and the standard allocated code amount for each block for each color component are obtained. Then, based on this information, the standard allocated code amount for each block for each color component is determined, and encoding processing is performed for each block for each color component so that it falls within the standard allocated code amount for each block for each color component. Implement.

Encoding processing is performed sequentially for each color component and in block order using the predicted quantization width for each color component, and while looking at the encoded output, the standard allocated code amount for each block and the block preceding the processing block are determined. The amount of code obtained by variable-length encoding is determined so that the amount of code generated in the processing block falls within the amount of allocated code for that block, which is obtained by adding up the amount of excess or deficiency (carryover amount) for the amount of allocated code in the amount of code generated in the processing block. Encoding is continued while monitoring the block, including EOB codes, and when the generated code amount or the allocated code amount is reached, the encoding of that block is terminated, or the DC component is always encoded even if it exceeds the allocated code amount. The amount of excess or deficiency with respect to the allocated code amount at that time is stored, and the next block is encoded.

Statistical processing predicts the optimal quantization width for each color component, optimally allocates the amount of code for each color component, and based on this, determines the standard allocated amount of code for each block for each color component. This is the only way to obtain statistics.

By using the optimized quantization width obtained through statistical processing in the next encoding process, it is possible to approach the desired total amount of code, or whether there is an upper limit on the amount of data for one image. In this case, the target code amount cannot be exceeded by even 1 bit, let alone 1 byte.

Therefore, control is performed based on the allocated code amount for each block. This is used for fine adjustment when the code amount during encoding exceeds the target code amount.The standard allocated code amount, which is the standard allocated code amount for each block, is determined. The amount of code generated by subtracting the amount of code generated from the amount of allocated code in the processing block immediately before the block is obtained from this standard allocated amount of code, and the remaining amount is obtained as a carryover amount. Except in special cases, encoding is controlled so that encoding is performed within the maximum frame for encoding processing.

Adjustment of the allocated code amount by this carryover differs depending on the code amount at the time of encoding, so in blocks where there is a surplus, the surplus is transferred to the block where there is a shortage, so that the overall target code is achieved. This is to rationally allocate the amount within the specified amount. This provides a function for fine adjustment when the standard code amount is exceeded. Each block is encoded sequentially from low frequency components to high frequency components, and when the guideline (allocated code amount per block) is exceeded, the encoding of any further high frequency components is discontinued.

This is because high frequency components of each block are omitted to minimize visual impact. Regarding the termination of the alternating current component, the fact that high frequency components tend to have less visual impact especially as the frequency becomes higher is utilized.

For blocks whose generated code amount including EOB is less than or exactly equal to the allocated code amount, encoding is completed without any problem, that is, EOB is output. In this way, if there is a surplus, it is added to the standard allocated code amount for the next block and is used as the allocated code amount for this block.

Encoding is aborted for a block that exceeds the limit on the way, and the encoding of that block is terminated, that is, an EOB is output.

This is not possible in reality unless the compression ratio is extremely high, but
In the unlikely event that the given allocated code amount is insufficient to encode even the DC component, only the DC component is encoded and the encoding of that block is completed, that is, an EOB is output. Then, the insufficient amount of code is carried over to the next program as a shortage, and the shortage is subtracted from the standard allocated code amount of that block to become the allocated code amount of that block. At this time, since EOB also has the sign -, EOB
The amount of code including the amount of code should be within the allocated code amount.

In this way, the purpose of the present invention is to prevent the allocated code amount from being wasted, and to carry over the surplus of the generated code amount to the allocated code amount for subsequent programs, and to By predicting the quantization width for each color component so that the ratio of statistics for each block is maintained when obtaining The code amount becomes more reasonable, and good results can be obtained for each color component.

The present invention basically employs a method called the 2-nozu method in which encoding is completed by performing two processes: "statistical processing" and "encoding processing." When quantizing and variable-length encoding this, the standard allocated code amount assigned to each block is used as a standard, and any variation in the code amount generated in each block is absorbed by the next block. The adjusted code amount is set as the allocated code amount, and block encoding is performed within this allocated code amount, and when the allocated code amount is exceeded, the encoding of that block is terminated. . However, image data is decomposed into DC and AC components by orthogonal transformation, and even if the code amount exceeds the allocated code amount for the DC component, which has a large visual impact, the coding will not be aborted and the next block will be processed. Subtract the allocated code amount by the excess amount,
Furthermore, if there is a risk that the amount of code will exceed the allocated code amount due to the alternating current component, the coding will be terminated. Further, when the amount of code generated by encoding is less than the allocated code amount, the surplus is added to the allocated code amount of the next block to increase the allocated code amount of the next block. In this way, when the amount of code in the previous processing block is over or under the allocated code amount, the excess or deficiency is reflected in the allocated code amount of the next block and adjusted.

In general, especially in the case of images with high or low spatial frequencies, the standard allocated code amount for each block is determined by proportionally allocating the total intended code amount to be stored per image according to the statistics of each block, If we adopt the conventional method of actually encoding using one quantization coefficient common to each predicted color component, the amount of allocated code may be insufficient for the luminance component block and redundant for the chrominance component block. Alternatively, the opposite may occur, and the encoding of color components with insufficient code amount is aborted midway, resulting in deterioration of image quality.However, according to the present invention, images with a wide range of spatial frequencies can be imaged. On the other hand, it becomes possible to perform encoding within a certain amount of code within a certain processing time using a simple circuit, without degrading the image quality as much as possible.

An embodiment of this device using the above principle will be described.

FIG. 1 shows an embodiment in which an image data encoding device according to the present invention is applied to a digital electronic still camera, and FIG. 2 shows a block diagram of the configuration of the image data encoding device according to the present invention. . Note that illustrations and explanations of mechanisms of the digital electronic still camera that are not directly related to the present invention will be omitted.

As shown in FIG. 1, a digital electronic still camera body (hereinafter referred to as electronic camera body) 1 includes an imaging system 40 that captures images, and performs predetermined signal processing on the output of this imaging system 40. The signal processing circuit 60 has functions such as orthogonal transformation, linear quantization, variable length coding, etc.
an encoding circuit 80 that compresses and encodes the output of 0 and outputs it;
The image data and quantization width (or information corresponding to this) encoded by this encoding circuit 80 are stored on a recording medium 7.
1, a switch 30 for setting and inputting a desired data compression ratio, and a control circuit 90 for controlling the entire system.

A switch 30 for setting an image compression rate is provided on the operating section of the electronic camera body 1, and the switch 30 is connected to a control circuit 90.

The imaging system 40 includes a lens 40a for forming an optical image.
and an image sensor 40b such as a CCD. The signal processing circuit 80 includes an amplifier 60 that performs amplification, noise removal, etc.
a and an A/D that converts analog signals into digital signals.
A converter 60b and a buffer memory 60 consisting of RAM, etc.
c, and a process circuit 80d that performs color signal formation and the like. The encoding circuit 80 includes, for example, an orthogonal transform circuit 4 that performs orthogonal transform such as DCT (discrete cosine transform), a quantization circuit 6 that performs linear quantization, and a Huffman encoding circuit 8 that performs Huffman encoding as variable length encoding. In addition, a quantization width prediction circuit 12, a code amount calculation circuit 14, and a code amount allocation circuit 2 are provided.
0. It has a code truncation circuit 16 and a control circuit 18 that performs control processing within the encoding circuit 80.

The recording system 70 includes an interface circuit 70a and a memory card 71 containing an IC memory used as a recording medium. The memory card 71 is removably attachable to the electronic camera body 1.

The control circuit 90 is realized by a microprocessor (MPtl).

FIG. 6 shows a perspective view of the external appearance of the electronic camera body 1. The figure shows a binocular type, and 48 is L in the operation section.
A CD (liquid crystal) display, 3o, is a switch 30 in the operation section, and is also provided with a tele/wide changeover switch, a shutter operation button 50, and the like. Also,
49 is a finder.

Under the control of the control circuit 90, the LCD display 48 displays various values and states such as the shooting mode, number of frames, date, and time. In this electronic camera, the image compression ratio can be set to a desired value by operating a switch 30 provided on the operating section of the electronic camera body 1. That is, the control circuit 9o is set in advance with standard multiple types of compression ratio information, and this is determined based on the value of the number of images that can be taken by operating the switch 30. The compression rate to be applied is determined from the capacity of the medium), and the determined compression rate value and the number of images that can be recorded in the memory cart are converted and displayed on the LCD display 48 of the operation unit. . When the user presses the switch 30, the control circuit 9o changes these values each time the switch 30 is pressed.

When the user stops pressing the switch 30 at the desired value while looking at the displayed change value, the control circuit 90 sets the command compression rate corresponding to the command value for the number of images that can be taken at that time. It has become. This is done by the control circuit 90 determining the standard total code amount per image determined according to the compression rate, and providing this to the encoding circuit 80 as target code amount setting information. In addition, when the shutter operation button 50, which is a trigger switch, is pressed,
When the shutter function is activated, a subject image is formed on the image sensor 40b, and a charge image is accumulated in the image sensor 40b corresponding to this image. can be obtained. These controls are also controlled by the control circuit 90.

The imaging system 40 in FIG. 1 includes a photographic lens 40a and a CC
It has an image pickup device 40b made of an image pickup device such as D, and converts an optical image formed on the image pickup device 40b by the photographic lens 40a into an image signal and outputs it to the signal processing circuit 60. The signal processing circuit 6° includes an amplifier 80a,
An A/D converter 60b1, a buffer memory 60c1, and a process circuit 8 (ld) are included. (abbreviated as Cr (chroma red)), B-Y (hereinafter, this B-
In addition to separating Y into each color component of Cb (abbreviated as chroma blue), gamma correction, white balance processing, etc. are performed.

The output video signal of the imaging system 40 that has been digitally converted by the A/D converter 60b is stored in a buffer memory [1iOc] having a capacity for one frame, and is read out and sent to the process circuit 60d. As a result, it is separated into a Y component, which is a luminance signal system, and Cr and Cb components, which are chroma (C: color difference signal) systems. For example, the image data stored in the buffer memory 60c is first processed by a process circuit to obtain Y component data of the image signal in order to perform statistical processing on the luminance signal, and this is then supplied to the encoding circuit 80. , Y component data is performed, and once this processing is completed, chroma system cr and Cb component data are processed and then encoded.

The signal processing circuit 60 has a blocking function, and the image data for the Y component and the Cr and Cb components (one frame or one field) is read out from the buffer memory 60c and processed. It is possible to perform blocking processing to divide the data (minutes) into blocks of a predetermined size. Here, as an example, the block size is 8×8 pixels, but this block size is not limited to 8×8, and the block sizes may be different for Y and C (chroma system).

In this embodiment, luminance system Y data is read out, divided into blocks, and given to the subsequent processing system to perform statistical processing on this Y component data.Once the statistical processing is completed, next chroma system Cr , In order to start statistical processing on the Cb acid component data, reading and blocking of the chroma-based Cr and Cb acid component data is started.

In the chroma system blocking, all of the Cr component image data is first blocked, and then the Cb acid component image data is blocked.

The encoding circuit 80 has a configuration shown in FIG. Second
In the figure, reference numeral 4 denotes an orthogonal transformation circuit, which receives input image data in blocks and performs two-dimensional orthogonal transformation on each block of this image data. As the orthogonal transformation, cosine transformation, sine transformation, Fourier transformation, Hadamard transformation, etc. can be used. By performing orthogonal transformation, image data as transformation coefficients can be obtained. Since the image data as the transform coefficient is orthogonally transformed, it is data obtained by decomposing the spatial frequency of the image into each frequency component.

6 is a quantization circuit, which receives the image data (transform coefficients) output from the orthogonal transform circuit 4, and in the first quantization, quantization width for each frequency component is set in advance according to each color component. The transform coefficients are quantized using the quantization width corrected by multiplying by the quantization coefficient α set in advance according to the shooting mode, and in the second time, the optimal quantization coefficient α determined by the previous processing is quantized. This is a configuration in which quantization is performed using

8 is a variable length encoding circuit, and the variable length encoding circuit 8 subjects the quantized output output from the quantization circuit 6 to variable length encoding. Huffman coding is used as variable length coding,
Use arithmetic encoding, etc. In variable length coding, the amount of code for each block, the amount of code for the entire image, etc. change for each image. Although the type of variable length encoding used is not directly related to the present invention, an example using Huffman encoding will be shown here.

The variable length encoding circuit 8 scans the input quantized transform coefficients from low frequency components to high frequency components by a method called zigzag scan in which the input quantized transform coefficients are scanned in the order shown in FIG.

The data of the first DC component [DC] in the scanning order in FIG. 7 is output by Huffman encoding the difference value from the DC component of the block that has been subjected to variable length encoding immediately before. AC component [A
For C3, look at the conversion coefficients in order from the 2nd to the 64th in the scanning order in Figure 7, and if the conversion coefficient is not 0 (
In other words, when a valid (valid) coefficient appears, it is encoded into two-dimensional Huffman code using the number of consecutive 0 (invalid) coefficients that existed immediately before it (zero run) and the value of that valid coefficient, and then output. do.

Further, if invalid coefficients continue after a certain coefficient up to the 64th coefficient, an EOB (end of block) code indicating the end of the program is output. Furthermore, when a truncation signal is input, encoding is terminated, and an EOB is added and output. Then, the code length generated for each code is output to the code amount calculation circuit 14.

The code amount calculation circuit 14 integrates the code length for each code of the inputted Y, Cr, and Cb color components, and calculates Y, Cr.

Collect the code amount data for each block of each Cb component, calculate the code amount for each color component, and the code amount for the entire image, and quantize the code amount data for each color component and the code amount for the entire image. The data is output to the width prediction circuit 12, and the data of the code amount for each color component and the code amount of the entire image is output to the code amount allocation circuit 20.

At the start of the first bus, the quantization width #j circuit 12 receives information on the target code amount from the control circuit 18, and uses this code amount information to calculate log (BRy +BRc) based on the equation (4) explained earlier. za x log
SF+b However, BR is the code amount (the number of bits per pixel), sp is the quantization coefficient, and a and b are constants.

The initial value of the quantization coefficient α is set using the relationship shown in FIG. ,
From the target edge code amount, which is the maximum allowable data amount per image, for example, using the relationships of equations (2) and (3), calculate the optimal color to get close to the target edge code amount by linear prediction. The quantization coefficient α for each component is predicted by taking into account the quantization coefficient actually used this time.

Here, how to set the quantization coefficient to an optimal value is
Since this is an important issue, I will explain this point a little.

It is well known that when image data is preprocessed, this output is quantized, and this quantized output is variable-length coded, the amount of code generated changes when the quantization width of this quantization is changed. . This means that variable length coding, typified by human coding, uses the bias in the probability of occurrence of data to be coded to reduce the amount of code required to represent that data. Therefore, since "changing the quantization width" also means changing the probability of occurrence of a quantized value, it can be seen that by changing the quantization width, the generated code amount also changes.

By the way, even if the same encoding is performed with the same quantization width, the amount of generated code will differ depending on the image data at that time. However, when the same encoding is performed on one image data by changing the quantization width, a certain relationship is obtained between the quantization width and the amount of generated code. Furthermore, it has become clear that when the relationship between the quantization width and the amount of generated code is determined for a large amount of image data, the relationship with the highest frequency of occurrence can be statistically obtained.

Specifically, in many cases, the following relationships were obtained.

That is, if the relative ratio to a certain quantization width is SP, and the amount of generated code is expressed as the number of bits per pixel (bi-sortorate), and this is BR, the relationship of equation (1) above holds true. In equation (1), if the encoding is the same, a is approximately constant regardless of the image, and b depends on the image.

The value of b has a certain distribution depending on the image, and a representative value can be obtained from this distribution of frequency of occurrence.

Therefore, the optimal quantization coefficient can be found from the relationship in equation (1).

The code amount allocation circuit 20 calculates each color from the code amount of the image data for each color component inputted from the code amount calculation circuit 14, the code amount for each block for each color component, the code amount of the entire image, and the target edge code amount. The allocated code amount for each component block is calculated and outputted to the encoding abort circuit 1B. The calculation method here is, for example, to proportionally allocate the target edge code amount based on the ratio of the code amount for each block for each color component. For example, by multiplying the code amount of a certain block by the target edge code amount and dividing it by the code amount of the entire image, the standard allocated code amount of the block is determined. Alternatively, the target edge code amount is proportionally distributed by the code amount for each color component, and this is proportionally distributed by the generated code amount for each block of that color component. As a result, the standard allocated code amount for each block is suppressed according to the actual code amount in that block, so that if the code amount is small, it will be kept in time, and if the code amount is small, it will be increased accordingly. Assigned.

The code amount allocation circuit 20 has a code amount information table and a block allocation code amount data table, and rewrites the code amount information of the corresponding block position in the code amount information table with the code amount information input from the code amount calculation circuit 14. , calculates the standard allocated code amount for each color component block from the code amount for each block and the code amount for the entire image input from the code amount calculation circuit 14, and the target edge code amount, and calculates the standard allocated code amount for each color component block. The data of the standard allocated code amount is stored in the block allocated code amount data table. The standard allocated code amount for each block in this block allocated code amount data table is given to the encoding abort circuit 16 when the corresponding block is subjected to variable length encoding processing.

The encoding abort circuit 16 adds an excess/deficiency amount to the reference allocated code amount of the corresponding block from the code amount allocation circuit 20, which is the difference between the generated code amount and the allocated code amount in the block processed one block before this block. is added as a carryover from the previous block to determine the allocated code amount for this block, and the code length of each block from the code amount allocation circuit 20 is subtracted from the allocated code amount, and the remaining allocated code amount is , when the amount of codes to be transmitted and the EOB code become smaller than the total code amount, an abort signal is output and the variable length encoding circuit 8
It has a function to complete the encoding of the block, subtract the transmitted code amount from the allocated code amount of the block, and hold that value as the remainder of the allocated code amount. Therefore, the encoding abort circuit 16 refers to this allocated code amount,
If the input code amount to be transmitted and the EOB code sent do not exceed the allocated code amount, the block is not truncated, the encoding of that block is finished, and the block should be sent from the allocated code amount. An operation is performed in which the amount of code is reduced and that value is held as the remainder of the allocated amount of code. In addition, the encoding abort circuit 16 refers to this allocated code amount, and if the allocated code amount is not sufficient to send out the input DC component code and EOB code, these two codes are After sending out the block, the coding of that block is finished, and the sent code amount is subtracted from the allocated code amount of the block. That is, the remainder of the allocated code amount is held as a negative value.

10 is a code output circuit, and this code output circuit 10 connects the variable length codes inputted from the variable length encoding circuit 8. The code output circuit 10 connects the variable length codes inputted from the variable length encoding circuit 8, and stores the connected codes in a recording medium such as a memory card. It functions to write to the recording system 22 that is stored in the memory.

This system first performs statistical processing (first bus) using the initial standard quantization coefficient α that is common to each color component determined according to the shooting mode, and then uses the information for each block that is necessary for optimization. The amount of information, the amount of information for each color component, the amount of information for the entire image, etc. are checked, and then processing begins to perform optimized encoding based on the information obtained through this statistical processing (second bus). . Therefore, we first block the image, quantize the elements of this blocked image using a standard quantization coefficient α common to each color component, and then variable-length encode the transform coefficients obtained by quantization. Then, from the code amount information of each element of each block obtained by this variable length encoding, the code amount for each color component, and the code amount information for the entire image, it is possible to obtain the optimal code amount for each color component. Predicting the necessary coding coefficient α, proportionally allocating the amount of code to be allocated to each color component, and determining the standard amount of allocated code for each block based on the allocated amount of code and the amount of code generated in each block for that color component. , transition to the processing mode of optimal encoding for the image to be processed based on these, blocking processing of the image in implementing this processing mode, prediction quantization coefficient α for each color component for the elements of this blocked image Quantization processing using , variable-length encoding of the transform coefficients obtained by this quantization, allocation of the standard allocated code amount for each block plus the carryover code amount from the processing block immediately before that block. In order to control the variable length encoding process within the code amount and to save the encoded information of the image to be processed, the procedure of outputting the variable length code code information obtained by controlling the variable length encoding process is performed. However, it is assumed that the entire control management is performed by the control circuit 18 in the figure. Incidentally, such a function of the control circuit 18 can be easily realized by using a microprocessor (MPU).

The above is the configuration of the encoding circuit 80.

The recording system 70 in FIG. 1 is an interface circuit 70a.
There is a recording medium 71 that is detachably connected to this.
The image data and quantization width (or information corresponding thereto) encoded and output by the encoding circuit 80 are recorded on the recording medium 71 via the interface circuit 70a.

Next, the operation of the present apparatus having the structure shown in FIGS. 1 and 2 will be described. In order to get an overall overview, the basic operation will first be explained with reference to FIG. 8, which is an operation transition diagram.

When using the camera, a camera user operates a switch 30 to set the desired number of images that can be taken. Thereby, the control circuit 90 determines the optimum code amount according to the set number of images that can be photographed, and provides this to the encoding circuit 80 as target code amount setting information. In this way, the number of images that can be taken is set.

When a photograph is taken next, the subject image is formed as an optical image on the image sensor 40b placed behind the photographic lens 40a. The image sensor 40b converts the formed optical image into an image signal and outputs the image signal.

The image signal obtained by the image sensor 40b is input to the signal processing circuit 60, where it is amplified by the amplifier circuit 60a in the signal processing circuit 60, A/D converted by the A/D converter 60b, and temporarily stored in the buffer memory 60c. Retained. Thereafter, it is read out from the buffer memory 80c, and subjected to band correction by the process circuit 60d in the signal processing circuit 60.
Processing such as color signal formation is performed.

Here, the subsequent processing of the first bus is Y (luminance component), Cr
, Cb (chroma system, both are color difference components) signals are performed in this order, so color signal formation is also performed in accordance with this order. That is, the image signal is read out in blocks of 8 x 8 pixel matrices, and the process circuit extracts Y component, Cr component (R-component), Cb acid component, B-Y component from this blocked image signal data. ), the signals of each color component are separated, and comma correction, white balance processing, etc. are performed.

The image signal data of each color component in the 8×8 matrix blocked image signal separated by the process circuit 60cl is input to the encoding circuit 80. As a result, image data for one frame (or field) is divided into blocks of the predetermined size and sequentially
The signal is input to the encoding circuit 80. In addition, the process circuit EiO
The image signals of each color component processed by cl are Y, C
Each color component of r and Cb is stored in a buffer memory,
It may be read and used in later processing.

In this embodiment, the signal processing circuit 60 outputs the Y component (luminance component) of the image signal data for one image, and after the subsequent processing (statistical processing) is completed, the Cr All of the component image data is divided into blocks, which are then subjected to statistical processing at a later stage.Then, the Cb acid component image is divided into blocks and which are subjected to statistical processing at a later stage. conduct.

The encoding circuit 80 supplies this input data received from the signal processing circuit 60 to the orthogonal transformation circuit 4. Then, the orthogonal transform circuit 4 performs two-dimensional orthogonal transform using, for example, discrete cosine transform (DCT) on each block of the input image data that has been divided into blocks (hereinafter referred to as block image data). This orthogonal transformation using DCT is a process of dividing a certain waveform into frequency components and expressing them with the same number of cosine waves of different frequencies as the number of input samples.

Then, the orthogonally transformed block image data (transformation coefficients) are stored at corresponding frequency component positions on an 8x8 matrix in a buffer memory (not shown) (the origin of the matrix is the DC component, and the rest are AC components and the origin is This is input into the quantization circuit 6.

Then, the quantization circuit 6 performs the first bus (first time) quantization on this block image data (transform coefficients).

In this first quantization, the image quality setting value set by the user at the time of shooting is applied to the quantization matrix for each preset frequency component (frequency components are determined corresponding to each matrix position of the block). The control circuit 18 corresponds to
The transform coefficients are quantized using the quantization width multiplied by the standard (provisional) quantization coefficient α given by (Figure 8 (hl,
D). The quantization matrices at this time may be the same for the luminance system and the chroma system, or better results may be obtained by setting quantization matrices suitable for each.

The quantized block image data (transform coefficients) is input to a variable length encoding circuit 8, where it is variable length encoded. The variable length encoding circuit 8 zigzags scans the input quantized transform coefficients in the order shown in FIG. 7, scanning from low frequency components to high frequency components. In other words, the conversion coefficients are stored in an 8 x 8 matrix corresponding to the frequency components, and since the one closest to the origin has a lower frequency, the zigzag scan scans from the lower frequency component to the higher frequency component. can.

Since the first data in the scanning order in FIG. 7 is the DC component DC, the data of this DC component DC is the DC component D of the block (previous block) that was subjected to variable length encoding immediately before.
The difference value djrf-DC with C is Huffman encoded (FIGS. 8(di) and (el)).

Regarding the alternating current component AC, look at the conversion coefficients in order from the 2nd to the 64th in the scanning order in Figure 7, and if the conversion coefficient is 0.
If a non-valid (that is, valid) coefficient appears, the number of consecutive 0 (invalid) coefficients that existed immediately before it (celloran)
Two-dimensional Huffman encoding is performed using and the value of its effective coefficient ((d2), (e2)).

Further, the variable length encoding circuit 8 provides an EOB (end of block) code indicating the end of the block when invalid coefficients continue after a certain coefficient up to the 64th coefficient. Then, the code length generated for that code is output to the code amount calculation circuit 14 (gl).

Such processing is executed for all blocks of one image.

When such processing for the Y component is completed, the same processing is performed for each of the C' and cb components.

On the other hand, the code amount calculation circuit 14 inputs Y and Cr.

In order to calculate the code amount for the entire image for each color component of each cb component, we calculated the code amount for each block of each Y, Cr, and Cb component, and calculated the code amount for each color component and the entire image. At the same time as performing integration (g2), the data of the code amount for each block is output to the code amount allocation circuit 20. The code amount allocation circuit 20 writes the code amount data for each block as the code amount information of the corresponding block position in the code amount information table.

Then, Y, Cr for all blocks for one image.

At the stage when all the Huffman encoding processing for each Cb component is completed, under the control of the control circuit 18, the code amount calculation circuit 4 predicts the quantization width using the data of the code amount for each color component and the code amount for the entire image. At the same time, the data of the code amount of the entire image is output to the code amount allocation circuit 20.

The quantization width prediction circuit 12 uses the input code amount for each color component, the code amount data of the entire image, and the target total code amount data to approximate the value of the target total code amount by linear prediction, for example. The optimal quantization coefficient α for each color component is predicted based on the quantization coefficient actually used (see Fig. 8 (h
2)).

Further, the code amount allocation circuit 20 calculates a standard allocated code amount for each block from the input code amount for each block, the code amount for the entire image, and the target total code amount, for example, by calculating the ratio of the code amount for each block. , is calculated by proportionally allocating the total target code amount (FIG. 8 (h3)).

Specifically, to determine the standard allocated code amount for a certain block, multiply the code amount of the block by the target total code amount, and divide it by the code amount of the entire image. This is the standard allocated code amount in a block.

Then, the calculated standard allocated code amount data for each block is stored in the block allocated code amount data table. The data of the allocated code amount for each block in this block allocated code amount data table is given to the encoding abort circuit 16 when the corresponding block is subjected to variable length encoding processing.

This completes the first pass, that is, the first encoding (statistical processing) for determining the allocated code amount for each block and optimizing the quantization width.

Next, the second pass processing begins. This second pass process is
It is a process to obtain a final encoded output optimized to fit within the target total code amount.

This treatment is first performed on the Y component, and after the Y component is completed, the Cr and Cb acid component treatments are performed. The Cr and Cb acid components may be treated in any order. Therefore, if the treatment order is Cr and Cb, the Cr component is treated first, and then the Cb acid component is treated.

If the order is Cb and Cr, the Cb acid component is treated first, and the Cr component is treated after the Cb acid component is finished.

In the second pass processing, the image data is first read out in blocks, and the Y component (luminance system) image signal data extracted from this block and output from the signal processing circuit 60 is input to the encoding circuit 80 (a ). The input blocked image data is input to the orthogonal transformation circuit 4 in the encoding circuit 801; and orthogonal transformation is performed again (b). The transform coefficients obtained by this orthogonal transform are input to the quantization circuit 6 and quantized again (e). However, the quantization coefficient α used at this time is determined by the quantization width prediction circuit 1 in the previous pass.
2 is the calculated optimal quantization coefficient α for prediction for each color component.

Next, the transform coefficients of the quantized block image data are input to the variable length encoding circuit 8. In variable length encoding, as in statistical processing, among the transform coefficients of this block image data, first the difference value difr-DC of the DC component DC is Huffman encoded ((di), (el)), and then the AC component Data is sequentially extracted from AC by zigzag scanning and two-dimensional Huffman encoding is performed ((d2)).

(e2)).

However, each time a Huffman code for one element (one position in the matrix) is generated, the code amount allocation circuit 20 sends the standard allocated code amount for that block stored in the allocated code amount data table to the encoding abort circuit. On the other hand, the encoding abort circuit 16 adds the standard allocated code amount of each block and the remainder of the allocated code amount in the encoding process of the previous block to obtain the allocated code amount of the corresponding block, and sends it out. If the allocated code amount is not exceeded even if the exponent code amount and the EOB code are sent, a process is performed to reduce the code amount to be sent from the allocated code amount of the block without generating an abort signal.

Then, when the total code amount of the code amount of the block to be transmitted and the EOB code exceeds the remaining code amount of the allocated code amount, the encoding abort circuit 16 sends the variable length encoder 8 to the variable length encoder circuit 8. A termination signal is output, Huffman encoding of the block is finished, and from the allocated code amount of the block,
The transmitted code amount is subtracted and the value is held as the remainder of the allocated code amount. Then, the variable length encoding circuit 8 moves on to Nofman encoding of the next block obtained from the quantization circuit 6.

In addition, in the encoding abort circuit 16, the DC component and the EOB
If the amount of allocated codes is insufficient to send out the codes, these codes are sent out and then sent to the coding abort circuit 16.
outputs an abort signal to the variable length encoding circuit 8 to terminate the Huffman encoding of the block, and subtracts the transmitted code amount from the allocated code amount of the block and holds the remainder of the allocated code amount as a negative value. do.

Therefore, the variable length encoding circuit 8 outputs the converted Huffman code to the code output circuit 10 until the truncation signal is input from the coding truncation circuit 16, and the Huffman code for all elements of the matrix is output before the truncation signal is generated. When the encoding is completed, the variable length encoding circuit 8 outputs the EOB code to the code output circuit 10. Furthermore, if the abort signal is input before the end of the 11-person encoding for all elements of the matrix, the variable length encoding circuit 8 outputs the BOB code to the code output circuit 10 instead of that code. become. The code output circuit 10 temporarily stores this encoded data.

The variable length encoding circuit 8 then proceeds to Huffman encoding of the next block obtained from the quantization circuit 6.

By repeating such operations and completing processing of all blocks of one screen image, all encoding processing is completed. Once the Y component is processed in this way, the chroma components (Cr, Cb) are processed in a similar manner. In processing this second component as well, the quantization circuit 6 uses the predicted optimal quantization coefficient α for the chroma component calculated by the quantization width prediction circuit 12 in the previous pass.

In this way, all the encoding processing is completed by completing the processing of the second bus for all blocks of the image for one screen for all color components.

At the end of this process, the encoder output circuit IO outputs the optimized Huffman encoded data for one image to the recording system 22, and stores it in a storage medium 7 such as a memory card in the recording system 22.
Write to 1 (f). This is done by the output of the code output circuit 10, which connects the Huffman codes, which are variable length codes from the variable length encoding circuit 8, and writes them by giving them to the memory cart, which is the storage medium 71. . A storage medium 7 based on the output of this code output circuit 10
Writing to I may be done all at once when the second bus is finished, but when the first bus is finished and the execution of the second bus begins, the result of stringing variable-length Huffman codes is written. , the data may be written to the storage medium sequentially in units of one byte or several bytes as soon as they are collected.

Note that, prior to this, the code output circuit 1 (l) writes information on the quantization matrix of each color component used for encoding into the header part of the stored data of the encoded image, and leaves it as a clue during playback.

As described above, in this device, statistical processing is first performed, and this statistical processing calculates the predicted value of the optimal quantization width for each color component and the statistical amount necessary for determining the standard allocated code amount for each block for each color component. Based on this information, we predict the optimal quantization width for each color component, allocate the amount of code for each color component, and calculate information on the allocated amount of code and the amount of generated code for each block for each color component. After determining the standard allocated code amount for each block for each color component using
For each color component, look at the encoded output for each block in that color component in order, and check the standard allocated code amount assigned to that block and the actual output code for the allocated code amount in the processing block immediately before that block. By aborting variable-length encoding so that the amount of excess or deficiency (carryover) is within the allocated code amount for that block, the amount of code for each block is controlled and the code amount is within the desired amount of code. This is an important point of the present invention.

Therefore, the block size, the type of orthogonal transformation, the type of variable length encoding, etc. used in this embodiment are not limited. Also, statistical processing does not necessarily have to be done once;
It may be omitted or may be repeated multiple times.

However, it is best to perform statistical processing once in terms of processing time and effects obtained. In addition, statistical processing does not necessarily require actual encoding to check the generated code amount; instead, it is a method that calculates the activity for each block and calculates the standard allocated code amount and quantization coefficient for each block from the results. It's okay.

In addition, it is not necessarily necessary to allocate the entire target total code amount to each block; for example, if the remaining amount when the target total code amount is allocated to the allocated code amount of each block is used for the first block, the previous processing block It is also possible to give it as an excess or deficiency.

Furthermore, in this embodiment, the EOB code amount was included in the determination of the standard allocated code amount for each block and the code amount control, but the EOB code amount does not change even if the quantization width changes. Of course, there is no problem in determining the standard allocated code amount for each block and controlling the code amount by excluding the EOB code amount in advance, but this is preferable since it improves accuracy. Similarly, in predicting the quantization width, it is preferable to perform prediction using a code amount excluding the EOB code amount.

Furthermore, the compression rate does not need to be variable; it may be fixed to one type of compression rate, and the target total code amount, provisional quantum coefficients, etc. may all be given as fixed values. The configuration will be simpler. Further, the image data buffer memory may be located between the orthogonal transform circuit 4 and the quantization circuit 6, and in this case, the blocking and orthogonal transform processes in the encoding process can be omitted.

However, in order to maintain accuracy, the size of the image memory becomes large in this case.

Also, process processing is performed before A/D conversion,
You may digitize it afterwards. In addition, in this device, variable-length encoding is performed for each block starting with low frequency components, and high frequency components, which have relatively little visual impact on image quality, are omitted by aborting the encoding, which improves image quality. Encoding can be performed with high compression while minimizing deterioration.

The present invention having the configuration shown in FIGS. 1 and 2 described above has the following features:
In short, based on the results of statistical processing, prediction of the quantization coefficient for each color component, proportional allocation of the allocated code amount for each color component based on the generated code amount for each color component, and the proportional allocation of the code amount and The standard code amount for each block for each color component is determined based on the generated code amount for each block for each color component, and the predicted quantization coefficient for each color component is used in the encoding process. The result is obtained by quantizing and variable-length encoding each block for each color component, adding the standard allocated code amount of the variable-length encoding target block and the excess or deficiency with respect to the allocated code amount of the processing block before that block. Variable-length encoding is performed while monitoring the code amount for each block so that the code amount resulting from the variable-length encoding is within the allocated code amount. When the code amount reaches the allocated code amount, the code amount for that block is However, even if the DC component exceeds the allocated code amount, it is always encoded, and the excess or deficiency of the allocated code amount at that time is memorized and the next block is encoded. It is something.

This is because if the allocated code amount for each block is simply calculated by multiplying the statistical amount of each block by proportionally allocating the target code amount, the allocated code amount for each block will be redundant for a certain color component. If there is a case where a certain color component is insufficient, the target total code amount is allocated based on the predicted ratio of the generated code amount for each color component, and then this is further allocated proportionally based on the statistics for each block. In other words, by determining the allocated code amount for each block, the ratio of the allocated code amount for each block for each color component and the ratio of the generated code amount for each color component during actual encoding can be made as close as possible. This means that the amount of code allocated to each block can be set to a reasonable value, and by making such a reasonable allocation in advance, balanced encoding can be achieved for each color component. This is to ensure that.

Therefore, according to the present invention, it becomes possible to encode images with a wide range of spatial frequencies within a certain amount of code within a certain processing time without degrading the image quality.

In compression encoding, the encoding circuit 80 having the configuration shown in FIG. The process is completed by two steps: finding the optimal quantization width for the first bus, executing the second bus using this optimal quantization width, and obtaining the final compressed encoded data. The optimal α for each color component and the standard allocated code amount for each block for each color component are determined.

FIG. 3 shows the processing flow of the encoding circuit 80.

In the 'ia3 diagram, in order to make it easier to understand the processing flow of the encoding circuit 80, the signal flow in the first bus is indicated by a dotted arrow ■, and the signal flow in the second path is indicated by a solid arrow ■. Each is illustrated. If we briefly follow the operation along the flow of this signal, we will see the following.

When encoding image data, a target total code amount is set in the control circuit 18 of the encoding circuit 80.

By operating the switch 30, the user sets the desired number of shots that can be taken, and the control circuit 90 selects the optimum code amount according to the set number of shots, and uses this as information on the target total code amount. This is realized by feeding the signal to the encoding circuit 80 as follows.

Note that, in the initial state, the number of images that can be taken is set to a predetermined standard number.

When photographing is performed, an image signal is output from the image sensor in the imaging system 40. This output image signal is converted into a digital signal in the signal processing circuit 60, stored in a buffer memory, and then read out in blocks of 8 x 8 pixels, starting with the Y component, then the Cr component, and then the cb
The acid component is separated.

This separation is first performed on the Y component, and the Y component image data is output in blocks of 8 x 8 pixels. The Y component image data is input to the orthogonal transform circuit 4, and orthogonally transforms each block (in this example, DCT; Jlt! Discrete cosine transform
e Co51ne Transform) is performed, and the data is decomposed into frequency divisions.

The DCT transform coefficients obtained by the orthogonal transform circuit 4 are input to the quantization circuit 6, while the target total code amount is output from the control circuit 18 to the quantization width prediction circuit 12.
2, the initial value of the quantization coefficient α is set from the target total code amount using the relationship of equation (1), and is output to the quantization circuit 6. The quantization circuit 6 linearly quantizes the transform coefficient using the inputted initial value quantization coefficient a (that is, provisional quantization coefficient α). The quantized transform coefficients are input to a variable length encoding circuit 8, where they are subjected to variable length encoding (in this example, 1 human encoding).

The quantization coefficients input here are scanned from low frequency components to high frequency components called zigzag scan, and the data of the first DC component is the DC component of the block that was variable-length encoded immediately before. The difference value is Huffman encoded and output.

For the AC component, look at the conversion coefficients in order from the 2nd to the 64th in the scanning order, and when a conversion coefficient that is not 0 (that is, valid) comes out, it is the consecutive O (zero; Two-dimensional Huffman encoding is performed using the number of invalid coefficients (zero run) and the values of their valid coefficients. Also, if invalid output continues after a certain coefficient up to the 64th coefficient, EOB indicating the end of the block will be displayed.
Outputs the sign of .

The variable length encoding circuit 8 performs encoding as described above,
Every time one code is generated, the code length is output to the code amount calculation circuit 14. The code amount calculation circuit 14 stores each input code length for each color component block, and also accumulates and stores the code lengths for each color component.

After completing the processing as in step 2 for the Y component, next
The same treatment is applied to the r component and the cb acid component.

When encoding for one image is completed, the code amount calculation circuit 14 adds the accumulated code amounts for each block for each color component and calculates the code amount for each color component and the code amount for the entire image as a total code amount value. do. The code amount for each color component and the total code amount value are output to the quantization width prediction circuit 12, and are also output to the code amount for each color component and the code amount allocation circuit 20 for the entire image.

When the above-described first pass encoding processing is completed, the code amount allocation circuit 20 receives the code amount for each color component inputted from the code amount calculation circuit 14, the code amount for the entire image, the target code amount, and the side number of each color component. Based on the code amount of the block, a standard allocated code amount for each block for each color component is calculated and stored by proportional allocation. In addition, the quantization width prediction circuit 12 calculates a more suitable quantization coefficient α for each color component from the picture image code amount obtained by encoding in the first pass and the target total code amount given from the control circuit 18. It is predicted each time and outputted to the quantization circuit 6.

Subsequently, a second pass of encoding processing is performed on the same image data. In the second pass, the image data read from the memory in the signal processing circuit 60 is separated into color components in the processing order in the first pass, and the image data of each component is
After processing such as 8-pixel blocking is performed, the data is input to the orthogonal transform circuit 4, and orthogonal transform (DCT transform) is performed for each block in order.
The T transform coefficients are input to the quantization circuit 6.

On the other hand, the quantization circuit 6 linearly quantizes the OCT transform coefficients using the corrected quantization width based on the new quantization coefficient α for each color component based on the given prediction. The quantized coefficients are input to the variable length encoding circuit 8, and the first
Huffman encoding is performed in the same manner as when encoding the path.

Here, the amount of code generated during encoding is obtained at the end of the first pass, and the amount of code to be encoded is determined from among the standard allocated code amount of each color component side number block stored in the code amount allocation circuit 20. The assigned code of the block is the sum of the standard assigned code amount of the block and the surplus/deficiency of the code amount (carryover amount) with respect to the assigned code amount of the encoding processing block immediately before the block to be encoded. Comparison is performed using the amount as a reference value, and when this reference value is exceeded, the subsequent coding within that block is aborted by the action of the code abort circuit 16. Furthermore, the excess or deficiency of the allocated code amount at the time when the encoding of the block is finished is stored in the code amount allocation circuit 20.

The encoded data controlled to the target total code amount by the above method is sequentially outputted to the recording system 70 via the code output circuit 1 (l) and recorded.

Next, reproduction of compressed and encoded recorded image data recorded on the recording medium 71 by the recording system 70 will be explained.

Figure 4 shows the configuration of the player. In the figure, 100 is a main body of the playback machine, and this main body 100 of the playback machine includes a reading section 102.
, a decoding circuit 104, a processing circuit 106, and a control circuit 108. The reading unit 102 can attach and detach the storage medium 71, and transfers the contents of the storage medium 71 to the interface circuit 110.
It is read out via .

The decoding circuit 104 has functional blocks as shown in FIG. That is, 112 is a Huffman decoding unit that decodes Huffman encoded data, and 114 is a Huffman decoding unit that decodes the data obtained by Huffman decoding based on the quantization width information read from the storage medium 71 and set and input. The inverse quantization unit 11B performs quantization, and the IDC performs inverse DCT transformation on the data obtained by inverse quantization and outputs it as video signal data.
T (inverse DCT transformation) unit, and 118 is a control unit that controls these.

The processing circuit 106 includes a buffer memory 120 and an encoder 1.
22 and D/A converter 124. The buffer memory 120 is a memory that temporarily holds the video signal data output from the decoding circuit 104, and the encoder 122 converts the video signal data read from the buffer memory 20 into an NTSC video signal.
The D/A converter 124 converts this NTSC video signal into analog and outputs it as a television video signal.

The control circuit 108 is in charge of overall control of the player main body 100, and controls the reading section 102 of the player main body 100 to read out information on the quantization width at the time of encoding.
As a result, the decoding circuit 104 sets the information on the quantization width at the time of encoding read from the recording medium 71 to the dequantization unit 114, and then the control circuit 108 causes the information recording medium 71 to be compressed and encoded. In order to read out the video signal data, control is performed to control the reading section lO'2. Further, although not shown, the main body 100 of the player has a frame forwarding switch and the like, and the video at the frame position designated by this switch can be played back. The control circuit 108 also performs such control.

Next, the operation of the reproducing machine having the above configuration will be explained. When the recording medium (memory card) 71 on which compression-encoded video signal data is recorded is attached to the reading section 102 of the playback main body 100, the control circuit 108 first instructs the reading section 102 to read the data at the time of encoding. Control is performed to read out information on the quantization width for each color component. As a result, the reading unit 102 reads the recording medium 71.
Information about the quantization width at the time of encoding is read from the quantization width, and this information is set in the inverse quantization unit 114 of the decoding circuit 104.

Next, the control circuit 108 controls the reading unit 102 to read the video signal from the recording medium 71, so the reading unit 102
sequentially reads video signals from the recording medium 71 and inputs them to the decoding circuit 104.

In the decoding circuit 104 that receives this, the Huffman decoding unit 1
12, the Huffman code is decoded to obtain quantized coefficients. The quantized coefficients obtained in this way are supplied to an inverse quantization circuit 114 for inverse quantization. The inverse quantization here is performed using the previously set information on the quantization width for each color component.

The transform coefficients obtained by inverse quantization are transferred to the IDCT section 11
6, each block is subjected to inverse DCT transformation and restored to the original video signal. In this way Y, Cr.

The video signal is restored in cbO order, output from the decoding circuit 104, and written to the buffer memory 120 in the processing circuit 106. When writing of the video signal data for one screen is completed, the video signal data is read out from the buffer memory 120 in the normal scanning order of the television signal, and the encoder 1
22, the signal is converted into an NTSC video signal. Furthermore, it is converted into an analog signal by the D/A converter 124,
Output. By inputting this video signal to a TV monitor, the image can be played back as a TV video and viewed as a video.Also, by feeding it to a printing device such as a video printer and printing it, a hard copy can be obtained, so it can be used as a photo, etc. I started to appreciate things in a similar way.

As explained above, the camera can set the desired number of shots that can be taken, and by setting the number of shots that can be taken, the camera automatically sets the corresponding compression rate, and the compression rate is initially determined according to this set compression rate. Using the provisional quantization width,
One screen's worth of photographed image data is quantized in turn for each color component in each block and variable length encoded, and from the resulting code amount of each color component and one screen's worth of photographed image data, the "optimum for each color component" is determined. Quantization width'' is predicted, and the allocatable code amount for each color component is determined by proportional allocation from the code amount for each color component, and the standard code amount to be allocated to each block for each color component is determined based on the code amount for each color component. Next, the captured image data for one screen is quantized again in order for each block of each color component using the predicted optimum quantization width for each color component. For variable length encoding of quantized output,
Variable-length code is used within this range, with the maximum allocated code amount being the standard allocated code amount for each block and the remainder of the allocated code amount during encoding in the encoding process of the block processed one before. By decoding the encoded video signal using the optimal quantization width used at the time of shooting, it is possible to encode at the desired compression rate without installing hardware for each compression rate. Similarly, image data encoded at a desired compression rate can be decoded using one common hardware without providing hardware for each compression rate.

As described above, in the embodiment, the encoding process is carried out on the first bus (statistical processing).
Although we used a two-bus method in which the process is completed in a total of two times, one time for the first pass and one time for the second pass (encoding process), it is not limited to this, and it will be even better if the first bus is repeated several times. results. Further, although an example is shown in which a memory card is used as the recording medium, other media such as a floppy disk, an optical disk, and a magnetic tape may also be used. Further, although the camera and the playback device are shown as separate bodies, it is also possible to use an integrated camera in which the camera also has the functions of the playback device. Furthermore,
In the embodiment, the quantization width value itself is recorded on the recording medium, but the quantization width value may be converted or encoded and then recorded. In addition, in the above embodiment, the quantization width was set based on the target total code amount, or in applications where multiple target total code amounts are switched and used by mote, the quantization width corresponding to each mode is set. Of course, it is also possible to prepare it in advance and use it by switching it depending on the mode.

Note that the present invention is not limited to the embodiments described above and shown in the drawings, and can be implemented with appropriate modifications within the scope of the gist.For example, the present invention is applicable not only to still images but also to moving images. This method can be applied to compression encoding of various images such as images.

In addition, in the above embodiment, information on the total code amount is given from the compression rate correspondence information, and a quantization coefficient for each color component that can give a quantization width corresponding to this total code amount is predicted. Quantization is performed using the quantization width based on the quantization coefficient, but the relationship between the quantization width and the compression rate correspondence information is calculated in advance and created in a table. This is stored in memory etc. and the compression It is also possible to output quantization width information (ie, quantization coefficient or quantization width value) directly from the rate correspondence information. In this way, when information corresponding to the desired compression ratio is input, information on the quantization width that is uniquely determined according to the input compression ratio information can be read and output, and optimal quantization can be performed immediately. Immediately after setting the width, the quantization circuit can quantize the image signal data using this quantization width.

According to this configuration, the information on the optimal quantization width corresponding to various compression ratios is stored in advance in a memory, etc., and the information on the optimal quantization width can be obtained by simply reading out the information in accordance with the input compression ratio correspondence information. Therefore, when encoding to fit within the target total code amount, the processing can be done in an extremely short time and the hardware can be simple.

〔Effect of the invention〕

As described in detail above, according to the present invention, it is possible to control the amount of code for images with a wide range of spatial frequencies without degrading the image quality, and to encode images within a certain amount of code within a certain processing time. Accordingly, it is possible to provide an image data encoding device and encoding method that are suitable for electronic camera systems and the like.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing a first embodiment of the present invention;
The figure is a block diagram showing an example of the main part configuration of the device of the present invention, FIG. 3 is a block diagram for explaining the flow of operation of the circuit shown in FIG. 2, and FIGS. 4 and 5 are block diagrams showing the configuration of the regenerator. 6 is a perspective view showing the external appearance of the electronic camera body according to the present invention, FIG. 7 is a diagram for explaining zigzag scanning of a block divided into 8×8 pixels, and FIG. 8 is a diagram showing the principle of the present invention. FIG. 9 is an operation transition diagram for explaining the conventional technique, and FIG. 1θ is a diagram for explaining the relationship between code amount and quantization width. 1... Electronic camera body, 6... Quantization circuit, 8...
・Variable length encoding circuit, 10... Code output circuit, 12.
... Quantization width prediction circuit, 14 ... Code amount calculation circuit, 1
6... Encoding abort circuit, 18.18a... Control circuit, 20... Code amount allocation circuit, 24... DCPM circuit, 30... Switch, 40... Imaging system, 48...・
LCD display, 60... Signal processing circuit, 80... Encoding circuit, 70... Recording system, 71... Recording medium, 9
0...Control circuit, 100...Player main body, 102.
Reading unit, 104...Decoding circuit, 106...Processing circuit, 108...Control circuit, 110...Interface circuit, 112...Huffman decoding unit, 114...Reverse seedling unit,
11B...IDCT (inverse DCT conversion) section, 118...
・Control unit. Applicant's agent Patent attorney Atsushi Tsuboi□−Shun0↓Fog BR φ and smoke bath

Claims (9)

[Claims] (1) Divide each of the color image data separated by color component into blocks of a predetermined number of pixels, perform orthogonal transformation for each divided block by color component, and then quantize the transformed output by quantization means. Then, this quantized output is given to a variable-length encoding means to be variable-length encoded, and when encoding image data, it is first quantized with a provisionally determined quantization width and then variable-length encoded. , performs statistical processing to obtain the code amount for each block, and then calculates the code amount for each block obtained through this statistical processing and the total code amount target value per image to find the appropriate amount to fit within this total code amount target value. The optimal quantization width and allocated code amount for each block are determined, and the transform output for each block is quantized according to the determined optimal quantization width, and the quantized output is allocated for each block. In an encoding device that performs a coding process of sequentially performing variable length coding while controlling the code amount for each block according to the code amount, each block is coded for each color component based on the variable length coding output during the statistical processing. a calculation means for calculating a code amount of the code amount; and predicting an optimal quantization width for each color component based on the total code amount target value, the calculated value of the calculation means, and the provisional quantization width used in the statistical processing. quantization width prediction means, and control means for controlling to perform quantization for each color component using the optimum quantization width for each color component predicted by the quantization width prediction means during the encoding process. An image data encoding device characterized in that: (2) Divide each of the color image data separated by color component into blocks of a predetermined number of pixels, perform orthogonal transformation for each divided block by color component, and then quantize the transform output by quantization means. Then, this quantized output is given to a variable-length encoding means to be variable-length encoded, and when encoding image data, it is first quantized with a provisionally determined quantization width and then variable-length encoded. , performs statistical processing to obtain the code amount for each block, and then calculates the code amount for each block obtained through this statistical processing and the total code amount target value per image to find the appropriate amount to fit within this total code amount target value. The optimal quantization width and allocated code amount for each block are determined, and the transform output for each block is quantized according to the determined optimal quantization width, and the quantized output is allocated for each block. In an encoding device that performs a coding process of sequentially performing variable length coding while controlling the code amount for each block according to the code amount, each block is coded for each color component based on the variable length coding output during the statistical processing. a calculation means for calculating a code amount of; and predicting an optimal quantization width for each color component based on the total code amount target value, a value calculated by the calculation means, and a provisional quantization width used in the statistical processing. quantization width prediction means for determining the reference code amount for each block for each color component by allocating the total code amount target value according to the value calculated by the calculation means; and determining means for determining the reference code amount for each block for each color component; Using the standard code amount of each block for each color component determined by the means, the code amount generated by encoding in the processed block processed before the next target block to be encoded among the blocks and the corresponding code amount. an allocation unit that allocates the code amount obtained by calculating the difference between the code amount allocated in the processed block and the code amount obtained by adding this difference to the reference code amount of the processing target block as the allocated code amount in the processing target block; terminating means for aborting the variable length encoding so as not to exceed the allocated code amount allocated by the allocating means in the encoding process of the target block; and prediction by the quantization width predicting means during the encoding process. An image data encoding device comprising: control means for performing control to perform quantization for each color component using the optimal quantization width for each color component. (3) Divide each of the color image data separated by color component into blocks of a predetermined number of pixels, perform orthogonal transformation for each divided block by color component, and then quantize the transformed output by quantization means. Then, this quantized output is given to a variable-length encoding means to be variable-length encoded, and when encoding image data, it is first quantized with a provisionally determined quantization width and then variable-length encoded. , performs statistical processing to obtain the code amount for each block, and then calculates the code amount for each block obtained through this statistical processing and the total code amount target value per image to find the appropriate amount to fit within this total code amount target value. The optimal quantization width and allocated code amount for each block are determined, and the transform output for each block is quantized according to the determined optimal quantization width, and the quantized output is allocated for each block. In an encoding device that performs sequential variable length coding while controlling the code amount for each block according to the code amount, the target compression rate is selected and the data should be stored according to the selected compression rate. Compression rate selection means for determining the target total code amount and the provisional quantization width; Calculation means for calculating the code amount for each block for each color component based on the variable length encoded output during the statistical processing; , quantization width prediction means for predicting the optimum quantization width for each color component based on the total code amount target value, the calculated value of the calculation means, and the provisional quantization width used in the statistical processing; and a control means for controlling quantization for each color component using the optimum quantization width for each color component predicted by the quantization width prediction means during encoding processing. Encoding device. (4) Divide each of the color image data separated by color component into blocks of a predetermined number of pixels, sequentially perform orthogonal transformation for each divided block by color component, and then quantize the transformed output by quantization means. Then, this quantized output is given to a variable-length encoding means to be variable-length encoded, and when encoding image data, it is first quantized with a provisionally determined quantization width and then variable-length encoded. , performs statistical processing to obtain the code amount for each block, and then calculates the code amount for each block obtained through this statistical processing and the total code amount target value per image to find the appropriate amount to fit within this total code amount target value. The optimal quantization width and allocated code amount for each block are determined, and the transform output for each block is quantized according to the determined optimal quantization width, and the quantized output is allocated for each block. In an encoding device that performs sequential variable length coding while controlling the code amount for each block according to the code amount, the target compression rate is selected and the data should be stored according to the selected compression rate. compression rate selection means for determining the target total code amount and the provisional quantization width; a calculation means for calculating the total code amount for each component and the total code amount for the entire image, based on the total code amount target value, the calculated value of the calculation means, and the provisional quantization width used for the statistical processing; quantization width prediction means for predicting an optimal quantization width for each color component, and distributing the total code amount target value according to the calculated total code amount for each color component; a determining means for determining a reference code amount for each block for each color component by distributing it according to the code amount; Among the blocks, the difference between the amount of code generated by encoding in the processed block processed before the block to be encoded next and the amount of allocated code in the processed block is calculated and this is used in the above processing. an allocation unit that allocates a code amount obtained by adding the code amount to the reference code amount of the target block as an allocated code amount in the target block; and an allocation unit that allocates the code amount obtained by adding the code amount to the reference code amount of the target block; and terminating means for aborting the variable length encoding so as not to exceed the code amount; 1. An image data encoding apparatus, comprising: a control means for controlling to perform image data encoding. (5) The quantization means performs quantization during the encoding process from data of low frequency components among the frequency component data obtained by the orthogonal transformation, and
The variable-length encoding means obtains the difference between the DC component and the DC component in the previous processing block among the frequency component data of the image data for each block obtained by the orthogonal transformation for each frequency component. , a DC component encoding processing unit that variable-length encodes this difference, and an AC component encoding processing unit that performs variable-length encoding on the AC component after the encoding processing unit finishes processing the DC component. 5. The image data encoding apparatus according to claim 1, wherein the encoding aborting means is provided only in the alternating current component encoding processing section. (6) Divide the color image data separated by color component into blocks each having a predetermined number of pixels, sequentially perform orthogonal transformation for each divided block, and then quantize the transformation output by a quantization means, and then, This quantized output is given to variable length encoding means to be variable length encoded, and when encoding data, each block is determined from the target total code amount to be stored and the statistics for each block determined in advance. When performing variable length encoding while controlling the code amount for each block according to the allocated code amount for each block, the quantization is performed using a provisional quantization width, and the quantized output obtained thereby is encoded. A first step of processing and checking the code amount of the entire image, the code amount for each color component, and the code amount for each block for each color component, and the code amount of the entire image obtained in this first step and the code amount for each color component. A second step in which the quantization width required for optimization is predicted for each color component from the code amount, and a standard allocation code for each color component and each block is calculated from the target total code amount and the code amount of each block for each color component. a third step of calculating the amount; a fourth step of performing the quantization for each block of each color component using the predicted quantization width of each color component; a fifth step of variable length encoding the quantized output of each block for each color component within the allocated code amount for that block; and a block to be currently variable length encoded in the fifth step. The difference between the code amount by variable length encoding in the block processed immediately before and the allocated code amount in this block is calculated, and this is added as a carryover amount to the standard allocated code amount in the block to be currently processed. and a sixth step of setting the amount of code to be allocated to the block to be currently processed. (7) Divide the color image data separated by color component into blocks each having a predetermined number of pixels, sequentially perform orthogonal transformation for each divided block, and then quantize the transform output by a quantization means, and then, This quantized output is given to variable length encoding means to be variable length encoded, and when encoding data, each block is determined from the target total code amount to be stored and the statistics for each block determined in advance. When performing variable length encoding while controlling the code amount for each block according to the allocated code amount for each block, by setting the compression rate, the target total code amount to be accommodated and a first step of determining a provisional quantization width, performing the quantization for each color component using the provisional quantization width, and encoding the resulting quantized output for each color component; A second step that examines the code amount for the entire image, the code amount for each color component, and the code amount for each block for each color component, and the optimal code amount from the code amount for the entire image and the code amount for each color component obtained in this second step. A third step of predicting the quantization width required for each color component, and calculating the standard allocated code amount for each block for each color component from the target total code amount and the code amount for each block for each color component. a fourth step; a fifth step of performing the quantization for each block for each color component using the predicted quantization width for each color component; and for each color component in this fifth step; a sixth step of variable-length encoding the quantized output of each block within the allocated code amount for that block; and processing immediately before the block to be currently subjected to variable-length encoding in the sixth step. The difference between the code amount by variable-length coding in the block that has been processed and the allocated code amount in this block is calculated, and this is added as the carryover amount to the standard allocated code amount in the block to be currently processed. A method for encoding image data, comprising: a sixth step of assigning a code amount to a block to be processed. (8) The image data encoding method according to claim (6) or (7), wherein the provisional quantization width uses a predetermined value depending on the total code amount per screen. (9) The image data code according to claim (6) or (7), wherein the provisional quantization width uses a predetermined value according to the selected total code amount per screen. method.