JP4253839B2 - Image compression method, image compression program, image compression apparatus, and digital camera - Google Patents

Description

  The present invention relates to an image compression method and an image compression apparatus suitable for use in an image photographing apparatus such as a digital camera or a video camera.

Conventionally, in an image capturing device such as a digital camera or a video camera, image quality setting is performed when a captured image is recorded, and compression using a quantization table of a value corresponding to the image quality setting is performed or the image quality setting is performed. Compression is performed by calculating a quantization table value corresponding to the corresponding file size.
As a method for determining the quantization table corresponding to the image quality setting, a technique for dividing the original image into a plurality of blocks, measuring the amount of compressed data for each divided block, and determining the quantization table for each block has been proposed. (For example, refer to Patent Document 1).

The quantization table selection method disclosed in Patent Document 1 will be described in detail. A DCT transform (discrete cosine transform) is generated for each divided block to generate transform coefficients, and the generated transform coefficients are converted into a plurality of quantization tables. Quantize with one quantization table selected from among them, compress the quantized transform coefficient by entropy coding, measure the amount of data after compression for each block, and set the result in advance When the amount of data after compression is larger than the target compressed data amount compared to the target compressed data amount per block, a quantization table with a high compression rate is selected, and in the opposite case, a quantum with a low compression rate is selected. Select the conversion table. Thereby, compression is performed using a quantization table suitable for the property of the image in units of blocks.
JP-A-8-204970 (6th page, FIG. 5)

However, the conventional image compression method described above has the following problems.
That is, when the quantization table is changed depending on the subject, there is a case where the image quality does not change much even when the file size increases, but the file size becomes small even though the image quality does not change much. Since there is no idea of selecting a quantization table, there is a problem that the image storage memory cannot be effectively used. Note that images that have little change in image quality even when the file size is large are images with many low-frequency components in the transform coefficient after discrete cosine transform, such as an image of a blue sky with no clouds.

  In addition, since the quantization table is determined for each block, there is a problem that it takes time for the compression process.

  The present invention has been made in view of such a conventional problem, and provides an image compression method and an image compression apparatus capable of effectively utilizing a memory for storing images and shortening the compression processing time of the image. With the goal.

In order to solve the above problem, an image compression method of the invention according to claim 1 comprises:
Divide the original image data into multiple blocks,
Each of the plurality of divided blocks is subjected to discrete cosine transform to obtain a transform coefficient,
A block having a high frequency component is detected as a block for calculating an image quality evaluation value from the conversion coefficients of the plurality of blocks obtained thereby,
Using the plurality of quantization tables having different values of one or more elements in the quantization table, the detected image quality evaluation value calculation block and the surrounding blocks adjacent to the image quality evaluation value calculation block are compressed. ,
The compressed image quality evaluation value calculation block and surrounding blocks adjacent to the image quality evaluation value calculation block are expanded using the quantization table used in the compression,
The pixel value of the peripheral portion of the image quality evaluation value calculation block after expansion and the peripheral pixel value of the peripheral block adjacent to the image quality evaluation value calculation block on the side in contact with the image quality evaluation value calculation block Using the difference between and the image quality evaluation value for each quantization table,
Obtain an approximate curve that approximates the relationship between each quantization table and the image quality evaluation value,
Based on the approximate curve, determine a quantization table number corresponding to the quantization table used for compression of the original image,
The original image data is compressed using a quantization table having the determined quantization table number.

  According to this method, since a good quantization table is selected using the image evaluation value and the original image is compressed, depending on the subject, the image quality may change even if the quantization table is changed and the file size is increased. It is possible to compress the non-existing files with the minimum file size. As a result, more images can be recorded in the image storage memory than before, and the image storage memory can be effectively used.

  In addition, an image quality evaluation value is obtained from a block for calculating an image quality evaluation value of a block having a high frequency component among the original image data divided into a plurality of blocks, and a good quantization table is selected using the obtained image quality evaluation value. Therefore, the image compression processing time can be greatly shortened as compared with the prior art that determines the quantization table for each block.

The image compression method of the invention according to claim 2
In the image compression method of the invention according to claim 1 ,
As a method of determining the quantization table number corresponding to the quantization table used for compression of the original image,
The differential value of the approximation curve corresponding to each quantization table is compared with a predetermined threshold value, among the quantization tables wherein the differential value is equal to or greater than a predetermined threshold value, decreases the value of each element of the quantization table is most The quantization table number is defined as a quantization table number corresponding to a quantization table used for compressing the original image.

  According to this method, a normalization table value capable of good compression can be calculated from the balance between image quality and data amount, and original image data can be compressed using the quantization table of the calculated normalization table value. That is, the image data can be compressed at the maximum compression rate without degrading the image quality.

The image compression program of the invention according to claim 3 is:
A block division process for dividing the original image data into a plurality of blocks;
DCT arithmetic processing for performing discrete cosine transform on each of the plurality of divided blocks;
An image quality evaluation value calculation block detection process for detecting a block having a high frequency component as a block for calculating an image quality evaluation value from the conversion coefficients of the plurality of blocks obtained by the discrete cosine transform;
Using a plurality of quantization tables having different values of one or more elements in the quantization table, each of the detected image quality evaluation value calculation block and each neighboring block adjacent to the image quality evaluation value calculation block A compression process for compressing each quantization table;
Decompression processing for decompressing each compressed image quality evaluation value calculation block and the image quality evaluation value calculation block using the same quantization table as in the compression,
The side in contact with the image quality evaluation value calculation block of the pixel value at the peripheral edge of the image quality evaluation value calculation block after expansion for each quantization table and the surrounding blocks adjacent to the image quality evaluation value calculation block An image quality evaluation value calculation process for calculating an image quality evaluation value for each quantization table using a difference from the pixel value of the peripheral edge of
An approximate curve calculation process for calculating an approximate curve that approximates the relationship between each quantization table and the image quality evaluation value;
Based on the calculated approximate curve, a quantization table number determination process for determining a quantization table number corresponding to a quantization table used for compression of the original image;
Original image data compression processing for compressing the original image data using a quantization table having the determined quantization table number;
Including
The image data of a subject photographed by an image photographing device by a computer is compressed.

  According to this program, the original image is compressed by selecting a good quantization table using the image evaluation value. Therefore, depending on the subject, the image quality may change even if the quantization table is changed and the file size is increased. It is possible to compress the non-existing files with the minimum file size. As a result, more images can be recorded in the image storage memory than before, and the image storage memory can be effectively used.

  In addition, an image quality evaluation value is obtained from a block for calculating an image quality evaluation value of a block having a high frequency component among the original image data divided into a plurality of blocks, and a good quantization table is selected using the obtained image quality evaluation value. Therefore, the image compression processing time can be greatly shortened as compared with the prior art that determines the quantization table for each block.

The image compression program of the invention according to claim 4 is:
As a method of determining the quantization table number corresponding to the quantization table used for compression of the original image,
The differential value of the approximate curve corresponding to each quantization table is compared with a predetermined threshold, and among the quantization tables in which the differential value is equal to or greater than the predetermined threshold, the value of each element of the quantization table is the smallest. The quantization table number is defined as a quantization table number corresponding to a quantization table used for compressing the original image .

According to this program, a normalization table value capable of good compression can be calculated from the balance between image quality and data amount, and original image data can be compressed using the quantization table of the calculated normalization table value. That is, the image data can be compressed at the maximum compression rate without degrading the image quality.

An image compression apparatus according to a ninth aspect of the present invention provides:
Block dividing means for dividing the original image data into a plurality of blocks;
DCT computing means for obtaining a transform coefficient by performing discrete cosine transform on each of the plurality of blocks divided by the block dividing means;
Image quality evaluation value calculation block detection means for detecting a block having a high frequency component from the conversion coefficient of each of the plurality of blocks obtained by the DCT calculation means, as an image quality evaluation value calculation block;
The image quality evaluation value calculation block detected by the image quality evaluation value calculation block detection means and the image quality evaluation value calculation using a plurality of quantization tables having different values of one or more elements in the quantization table Compressing the surrounding blocks adjacent to the image block, and further compressing the compressed image quality evaluation value calculating block and the surrounding blocks adjacent to the image quality evaluation value calculating block used in the compression Compression / expansion means for expanding using
The pixel value of the peripheral portion of the image quality evaluation value calculation block after expansion and the peripheral pixel value of the peripheral block adjacent to the image quality evaluation value calculation block on the side in contact with the image quality evaluation value calculation block Image quality evaluation value calculation means for obtaining an image quality evaluation value for each quantization table using the difference between
An approximate curve calculation means for obtaining an approximate curve that approximates a graph representing the relationship between each quantization table and the image evaluation value;
A quantization table number determining unit that determines a quantization table number corresponding to a quantization table used for compression of the original image based on the approximate curve obtained by the approximate curve calculating unit;
Original image data compression means for compressing the original image data using the quantization table of the quantization table number determined by the quantization table number determination means;
It is characterized by comprising.

  According to this configuration, since a good quantization table is selected using the image evaluation value and the original image is compressed, depending on the subject, the image quality may change even if the quantization table is changed and the file size increases. It is possible to compress the non-existing files with the minimum file size. As a result, more images can be recorded in the image storage memory than before, and the image storage memory can be effectively used.

  In addition, an image quality evaluation value is obtained from a block for calculating an image quality evaluation value of a block having a high frequency component among the original image data divided into a plurality of blocks, and a good quantization table is selected using the obtained image quality evaluation value. Therefore, the image compression processing time can be greatly shortened as compared with the prior art that determines the quantization table for each block.

According to an eleventh aspect of the present invention, an image compression apparatus includes:
In the image compression apparatus of the invention according to any one of claims 8 to 10,
The quantization table number determining means includes
Comparing the derivative of the approximate curve corresponding to each quantization table with a predetermined threshold;
Among the quantization tables in which the differential value takes a value within a predetermined range, the quantization table number where the value of each element of the quantization table is the largest corresponds to the quantization table used for compression of the original image. The quantization table number is determined.

  According to this configuration, a normalization table value capable of good compression can be calculated from the balance between image quality and data amount, and the original image data can be compressed using the quantization table of the calculated normalization table value. That is, the image data can be compressed at the maximum compression rate without degrading the image quality.

The digital camera of the invention according to claim 7 is:
The image compression apparatus according to claim 5 or 6 is provided.

  According to this configuration, it is possible to provide a digital camera in which the image compression processing time is short and the image storage memory can be used effectively.

  In the present invention, since a good quantization table is selected using the image evaluation value and the original image is compressed, there is no change in image quality even if the quantization table is changed and the file size is increased depending on the subject. Can be compressed with a minimum file size. As a result, more images can be recorded in the image storage memory than before, and the image storage memory can be effectively used.

  In addition, an image quality evaluation value is obtained from a block for calculating an image quality evaluation value of a block having a high frequency component among the original image data divided into a plurality of blocks, and a good quantization table is selected using the obtained image quality evaluation value. Therefore, the image compression processing time can be greatly shortened as compared with the prior art that determines the quantization table for each block.

  Also, since a normalization table value that can be favorably compressed can be calculated from the balance between the image quality and the data amount, the original image data can be compressed using the quantization table of the calculated normalization table value. That is, the image data can be compressed at the maximum compression rate without degrading the image quality.

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.
(Embodiment 1)
FIG. 1 is a functional block diagram of an image compression apparatus according to Embodiment 1 of the present invention.
In FIG. 1, an image compression apparatus according to the present embodiment is equipped with a microcomputer (not shown) and compresses input image data by software control. The microcomputer includes an image data acquisition unit 9 and a block. Division unit 10, DCT operation unit 11, image quality evaluation block detection unit 12, quantization table unit 13, quantization unit 14, entropy encoding unit 15, block input unit 16, entropy inverse encoding unit 17, inverse quantization unit 18, an inverse DCT calculation unit 19, a quantization table selection unit 20, an image quality evaluation value calculation unit 21, a threshold adjustment unit 22, a data output unit 23, and the like. The quantization table unit 13 has a plurality of quantization tables, which are stored in a nonvolatile memory (not shown) such as an EEPROM (Electrically Erasable and Programmable Read Only Memory). Needless to say, the microcomputer has a memory storing a program for performing image compression processing, and performs image compression processing described below in accordance with the program.

The image data acquisition unit 9 acquires original image data input from the outside, and inputs the acquired original image data to the block division unit 10. The block division unit 10 divides the original image data input from the image data acquisition unit 9 into a plurality of blocks and inputs the blocks to the DCT calculation unit 11 for each block. The DCT calculation unit 11 performs DCT conversion on the multi-value image for each block input from the block division unit 10 to obtain a conversion coefficient (referred to as a coefficient obtained by performing DCT conversion). Thereby, the conversion coefficient of the original image data divided into a plurality of blocks is obtained. Here, the DCT conversion is to convert image data represented by a digital signal into frequency components.
Then, the image quality evaluation block detection unit 12 detects an “image quality evaluation value calculation block” including a large amount of high frequency components from the conversion coefficients of each of the plurality of blocks obtained by the DCT calculation unit 11 when calculating the image quality evaluation value. Then, the conversion coefficient of the detected “image quality evaluation value calculation block” and the conversion coefficients of the four neighboring blocks adjacent to the “image quality evaluation value calculation block” are input to the quantization unit 14. In selecting a block with many high frequency components, for example, it can be performed by the following method. That is, first, for each block, the sum of squares of conversion coefficients of the “high frequency component” part and the sum of absolute values are obtained. Here, the range of the conversion coefficient representing the “high frequency component” is determined in advance, for example, using 4 × 4 conversion coefficients in the lower right, and the sum of the squares of the conversion coefficients in the range or the sum of the absolute values is determined. Try to ask. Then, a block that is determined by a method such as a block having the largest value, a plurality of blocks having this value equal to or greater than a predetermined value, or a predetermined number of blocks having a larger value. The image quality evaluation value calculation block is selected.

  The quantization table unit 13 has n (for example, 256) quantization tables having different quantization data (the values of the elements of the quantization table; the same applies hereinafter) and compresses the original image. The quantization table specified by the quantization table selection unit 20 is supplied to the quantization unit 14 and the inverse quantization unit 18 when the original image after compression is expanded. The n quantization tables included in the quantization table unit 13 are numerical values ((1), (2),..., (N) for designating the quantization table as shown in FIG. The value of the quantized data increases as the "quantization table number") becomes younger. In this case, the greater the value of the quantized data, the higher the compression ratio and the lower the image quality. Further, the file size when the original image is compressed using each of the n quantization tables as shown in FIG. 2 becomes smaller as the value of the quantized data becomes larger as shown in FIG. Here, FIG. 3 is a graph showing the quantization table number on the horizontal axis and the file size after compressing the original image on the vertical axis. Note that the compression ratio of the original image with respect to the value of the quantization table number is not necessarily a proportional graph as shown in FIG. 3, but is approximately proportional to a standard image. It is desirable to use a quantization table that can be assumed to be present.

  Returning to FIG. 1, when calculating the image quality evaluation value, the quantization unit 14 converts the conversion coefficient of the “image quality evaluation value calculation block” detected by the image quality evaluation block detection unit 12 and the “image quality evaluation value calculation block”. Quantize the transform coefficients of neighboring blocks adjacent to the block using the quantization table given from the quantization table unit 13. Further, at the time of image compression after calculating the image quality evaluation value, the transform coefficient of each of the plurality of blocks obtained by the DCT calculation unit 11 is quantized using the quantization table given from the quantization table unit 13. To do.

The entropy encoding unit 15 performs entropy encoding on the transform coefficient quantized by the quantization unit 14, compresses it, and outputs it to the block input unit 16. The block input unit 16 takes in the compressed data output from the entropy encoding unit 15 and inputs the compressed data to the entropy inverse encoding unit 17. The entropy inverse encoding unit 17 performs entropy inverse encoding on the compressed data input from the block input unit 16 using Huffman encoding or the like to obtain quantized transform coefficients. Then, the obtained quantized transform coefficient is input to the inverse quantization unit 18. The inverse quantization unit 18 inversely quantizes the quantization transform coefficient input from the entropy inverse encoding unit 17 using the quantization table (quantization table used at the time of compression) given from the quantization table unit 13. Get the conversion factor. Then, the obtained transform coefficient is output to the inverse DCT transform unit 19. The inverse DCT calculation unit 19 performs inverse DCT transform on the transform coefficient restored by the inverse quantization unit 18 to obtain a pixel value. That is, the transform coefficient restored by the inverse quantization unit 18 is subjected to inverse DCT transform to obtain a pixel value (hereinafter, this pixel value is referred to as “restored pixel value”). At the time of calculating the image quality evaluation value, only the transform coefficient of the image quality evaluation value calculation block restored by the inverse quantization unit 18 and the transform coefficients of the four neighboring blocks adjacent to the image quality evaluation value calculation block are subjected to inverse DCT conversion. If each pixel value is restored, it is more preferable in that the amount of calculation can be reduced. The obtained restored pixel value is output to the image quality evaluation value calculation unit 21.
The image quality evaluation value calculation unit 21 calculates an image quality evaluation value based on the output restored pixel value, and inputs the calculated image quality evaluation value to the quantization table selection unit 20.

Here, an example of a method of calculating the image quality evaluation value will be described with reference to FIG.
Various methods can be used as the image quality evaluation value calculation method. Here, “image quality evaluation value calculation block” that contains a lot of high-frequency components as the image is compressed by the quantization table. An example of an image quality evaluation method focusing on the feature that the high-frequency component is lost and the pixel value of the restored image becomes discontinuous at the boundary between adjacent blocks will be described.

The image quality evaluation value calculation unit 21 first obtains the restored pixel values of the image quality evaluation value calculation block and each of the four neighboring blocks adjacent thereto. Then, restoring the pixel values of the peripheral portion of the image quality evaluation value calculation block BKh as shown in FIG. _{4 (A 11, ..., A} 81, ..., A 88, ..., A 18, ..., A 11) and, around Among the pixels of each of the four adjacent blocks BKn1 to BKn4, the restored pixel values (B ₁₈ ,..., B ₈₈ , E ₁₁ ,..., E ₁₈ , D on the side in contact with the image quality evaluation value calculation block _{81 ..., D 11, C 88} , ..., obtaining the difference between _{C 81).} Then, as shown in Expression (1), the sum of differences is obtained to obtain an image quality evaluation value. In addition, although the example which uses the square sum of a difference value is shown in Formula 1, it is good also as an image quality evaluation value by calculating | requiring the absolute value of each difference value, and calculating | requiring the sum of this absolute value.

  Returning to FIG. 1, the quantization table selection unit 20 instructs the quantization table unit 13 to provide the quantization table 14 and the inverse quantization unit 18 with the quantization table specified by itself when calculating the image quality evaluation value. For the quantization unit 14, the image quality evaluation value calculation block BKh and the neighboring blocks BKn1 to BKn4 adjacent to the image quality evaluation value calculation block BKh using the quantization table given from the quantization table unit 13 are used. The image quality evaluation value calculation block BKh and the image quality evaluation value calculation block compressed using the quantization table given from the quantization table unit 13 are instructed to be compressed. Instruct to extend each of the neighboring blocks BKn1 to BKn4 adjacent to BKh. Then, after issuing these instructions, when the image quality evaluation value calculation unit 21 calculates the image quality evaluation value, the value is stored.

  This process is performed for each of a plurality of quantization tables having different quantization table numbers, and the image quality evaluation value obtained in each is stored. As a plurality of quantization tables to be designated, for example, as shown in FIG. 3, a quantization table corresponding to an arbitrary plurality of locations (for example, 4 points) from 256 quantization tables is used. When the image quality evaluation value for the image quality evaluation value calculation block BKh is calculated from each of these four quantization tables, quantization is performed using an approximation method such as a cubic spline function based on these image quality evaluation values. An approximate curve of a graph representing a change in image quality evaluation value when the table number is changed is calculated.

  After calculating the approximate curve, the differential value of the approximate curve in each quantization table number is compared with a predetermined threshold, and the quantization table with the largest quantization table number among the quantization tables whose differential value is equal to or greater than the predetermined threshold That is, the lowest compression rate is detected, and a value larger by “1” than the quantization table number is set as a quantization table number to be selected by the quantization table selection unit 11. That is, the quantization table number of the quantization table that can reduce the file size without degrading the image quality is calculated from the balance between the image quality and the data amount. Then, the quantization table having the calculated quantization table number is designated in the quantization table unit 13, and the original image data is converted to the quantization unit 14 using the quantization table designated in the quantization table unit 13. Give instructions to compress.

  Here, FIG. 5 is an example of the approximate curve of the image quality evaluation value versus the normalization table when the reciprocal of the image quality evaluation value is taken. Here, the reciprocal is taken in the example description of the image quality evaluation value calculation method described above, and the pixel of the restored image at the boundary with the adjacent block as the image is compressed by the quantization table. This is because an image quality evaluation value using a feature in which values become discontinuous is used, and a curve that takes a large value when the image quality is high is used. Image quality evaluation values other than the selected four-point quantization table can be obtained from this approximate curve. The quantization table selection unit 20 compares the differential value of the approximate curve with a predetermined threshold value, and among the quantization tables whose differential value is equal to or greater than the predetermined threshold value, the one with the highest quantization table number, that is, the most compressed A value having a low rate is detected, and a value larger by “1” than the quantization table number is set as a quantization table number to be selected by the quantization table selection unit 20. That is, when an image is compressed using a quantization table having a quantization table number 1 less than the quantization table as compared with the case where the image is compressed with a certain quantization table, the image compression rate is In this embodiment, the degree of decrease in the image quality evaluation value is determined based on the differential value of the image quality evaluation value. Then, when the degree of decrease in the image quality evaluation value is greater than or equal to a predetermined value, that is, when the quantization table number is changed by 1, a quantization table number of a quantization table that causes image quality deterioration greater than or equal to a predetermined value is obtained. A quantization table having a quantization table number that is “1” larger than the quantization table number of a limit table that does not cause image quality deterioration, that is, a quantization table that causes image quality deterioration of a predetermined level or more is defined as a current processing target. The quantization table is selected as being suitable for compression of the current image. In the quantization table selection unit 20, the threshold value to be compared with the differential value of the approximate curve can be adjusted. The threshold adjustment unit 22 adjusts the threshold of the quantization table selection unit 20, and the threshold can be set to an arbitrary value by operating this.

  The data output unit 23 outputs the original image data compressed by this apparatus and the decompressed original image data to the outside. In this case, the decompressed data can be output because, for example, the compressed data of the original image that has been compressed and stored by the present apparatus is converted into the entropy inverse encoding unit 17, the inverse quantization unit 19, and the inverse DCT operation unit 19. This is so that it can be restored using. Since this apparatus can compress and decompress original images, it can be easily applied to digital cameras and video cameras.

Next, the operation of the image compression apparatus having the above configuration will be described.
When the original image data is acquired, it is divided into a plurality of blocks, and a multi-valued image for each block is DCT transformed to obtain a frequency distribution. A block having a high frequency component in the frequency distribution is detected as an image quality evaluation value calculation block BKh. Here, a plurality of blocks may be used as the image quality evaluation value calculation block. Then, with respect to the detected image quality evaluation value calculation block BKh and the neighboring blocks BKn1 to BKn4 adjacent to the detected block, a predetermined number (for example, four) of quantization tables having different quantized data values are used. Compress / decompress each quantization table.

  And the pixel value of the peripheral part of the image quality evaluation value calculation block BKh after decompression, and the pixel value of the peripheral part on the side in contact with each of the neighboring blocks BKn1 to BKn4 adjacent to the image quality evaluation value calculation block BKh Is obtained, and the obtained sum is used as an image quality evaluation value. Then, based on the image quality evaluation value for the image quality evaluation value calculation block BKh for each quantization table, the change in the image quality evaluation value when the quantization table number is changed using an approximation method such as a cubic spline function An approximate curve of the graph representing is calculated. Then, the calculated derivative value of the approximate curve is compared with a predetermined threshold value, and the quantization table whose slope is equal to or greater than the predetermined threshold value has the largest quantization table number, that is, the one with the lowest compression rate. It is detected and a value larger by “1” than the quantization table number is set as a quantization table number to be selected by the quantization table selection unit 11. Then, the original image data is compressed using a quantization table having the calculated quantization table number.

  As described above, according to the image compression apparatus according to the present embodiment, since the original image data is compressed by selecting a good quantization table using the image quality evaluation value, the file size is changed by changing the quantization table. It is possible to compress an image in which image quality degradation hardly occurs even when the image size becomes small, with a minimum file size without causing significant image quality degradation. As a result, more images can be recorded in the image storage memory than before, and the image storage memory can be effectively used.

  Further, an image quality evaluation value is obtained from the image quality evaluation value calculation block BKh for a block having a high frequency component among the original image data divided into a plurality of blocks, and a good quantization table is selected using the obtained image quality evaluation value. Therefore, the image compression processing time can be greatly shortened as compared with the prior art that determines the quantization table for each block.

  In the above embodiment, image evaluation values are obtained for the four selected quantization tables, and an approximation of the image quality evaluation value versus the normalization table using a cubic spline function based on the obtained four image evaluation values. Although the curve is calculated, it is only necessary to calculate a smooth curve that passes (or approximates) the four image evaluation values. Therefore, the interpolation method is not limited to the cubic spline function, and other interpolation methods may be used.

  Further, in the above embodiment, the point where good compression is possible is obtained from the slope of the approximate curve, but the value of the image quality evaluation value itself may be compared with a predetermined threshold value. In this method, there is a tradeoff with the dynamic range of the image quality evaluation value (how much the value changes), and as an improvement, for example, the image evaluation value has been reduced to, for example, half between the maximum value and the minimum value. It is also possible to use a method in which the point is lowered to a predetermined ratio.

  Further, in the above embodiment, the original image data is divided into a plurality of blocks, the frequency distribution for each block is examined, the image quality evaluation value is obtained based on the block having many high frequency components, and the obtained image quality evaluation value is used. Although an original image is compressed by selecting a good quantization table, various methods for obtaining an image quality evaluation value have been proposed in the past, and they may be used. As a conventional method for obtaining an image quality evaluation value, for example, Japanese Patent Application Laid-Open No. 10-200493, Japanese Patent Application Laid-Open No. 8-205156, Japanese Patent Application Laid-Open No. 8-195888, Japanese Patent Application Laid-Open No. 7-184062, and Japanese Patent Application Laid-Open No. 6-334988. There are various inventions described in Japanese Patent Laid-Open No. 6-125545.

(Embodiment 2)
FIG. 6 is a block diagram showing a schematic configuration of a digital camera according to Embodiment 2 of the present invention.
In FIG. 6, a digital camera 30 according to the present embodiment includes a CCD (Charge Coupled Device) 31, a DSP (Digital Signal Processor) / CPU (Central Processing Unit) 32, a TG (Timing Generator) 33, and a unit circuit. 34, a display unit 35, a key input unit 36, a voice processing unit 37, an address / data bus 38, a DRAM (Dynamic Random Access Memory) 39, a built-in flash memory 40, and a card I / F (interface). 41 and a memory card 42.

  The DSP / CPU 32 is a one-chip microcomputer that executes various digital signal processing functions including image data compression / decompression and audio data processing, and also executes a program having a function of controlling each part of the digital camera 30. In particular, the DSP / CPU 32 has a function of compressing and storing captured images with priority on image quality, and is executed by selecting an image quality priority compression mode. In this case, the image quality priority compression process is obtained by programming the same process as the image compression process of the image compression apparatus according to the first embodiment described above.

  A TG (Timing Generator) 33 that drives the CCD 31 is connected to the DSP / CPU 32, and a unit circuit 34 to which an analog imaging signal corresponding to the optical image of the subject output from the CCD 31 is input to the TG 33. It is connected. The unit circuit 34 is a CDS (Correlated Double Sampling) circuit for sampling and holding the image pickup signal output from the CCD 31 and holding it, a gain adjustment amplifier (AGC) for amplifying the image pickup signal, and the amplified image pickup signal digitally. It is composed of an A / D converter (AD) for converting it into a signal, and the output signal of the CCD 31 is sent to the DSP / CPU 32 through the unit circuit 34 as a digital signal.

  The DSP / CPU 32 is connected to a display unit 35, a key input unit 36, and a voice processing unit 37, and is also connected to a DRAM 39, a built-in flash memory 40, and a card I / F 41 via an address / data bus 38. The card I / F 41 is connected to a removable memory card 42 attached to a throttle (not shown) of the main body of the digital camera 30.

  The display unit 35 includes a color LCD (Liquid Crystal Display) and its driving circuit, and displays a subject image captured by the CCD 31 as a through image when in a shooting standby state, and a memory serving as a storage memory when reproducing a recorded image. Display the recorded image read from the card 42 and expanded. The key input unit 36 includes a plurality of operation keys such as a shutter button (half press, full press), a recording start / end button used for moving image shooting, a power key, a MENU key, and the like according to the key operation by the user. The signal is output to the DSP / CPU 32.

  The sound processing unit 37 includes a built-in microphone, an amplifier, an A / D converter, a built-in speaker, a D / A converter, and the like. When shooting a still image or a moving image with sound, the sound input to the built-in microphone is converted into a digital signal. The data is converted and sent to the DSP / CPU 32. The audio data sent to the DSP / CPU 32 is sequentially accumulated in the DRAM 39 and finally stored in the memory card 42 together with the image data generated by the DSP / CPU 32. The audio processing unit 37 reproduces audio based on audio data attached to each video and outputs the audio from a built-in speaker (not shown) when reproducing a still image or moving image with audio. In addition, various notification sounds are reported by a built-in speaker as necessary.

  The DRAM 39 is also used as a buffer memory for temporarily storing digitized subject image data and the like as a working memory of the DSP / CPU 32 after being imaged by the CCD 31, and a moving image file recorded on the memory card 42 is stored in the card. The data is stored via the I / F 41 and the address / data bus 48, and processing necessary for editing work is performed.

  The built-in flash memory 30 executes various control modes required for control of each unit by the DSP / CPU 22, that is, various modes given to the digital camera such as a still image shooting mode, a moving image shooting mode, and an image quality priority compression mode to be described later. And a threshold value used in the image quality priority compression mode (see step S29 in FIG. 10) are stored, and the DSP / CPU 22 operates according to the program.

Next, the operation of the digital camera 30 configured as described above will be described.
"Processing procedure in still image shooting mode"
FIG. 7 is a flowchart showing the processing procedure of the DSP / CPU 32 when the user sets the still image shooting mode.
When the user operates the key input unit 36 to set to the still image shooting mode, the digital camera 30 can capture the scenery and scenery that he / she wants to shoot, and the captured video can be recorded on a memory card or the like. .

  That is, when the still image shooting mode is set, the DSP / CPU 32 starts shooting by the CCD 31 and displays a through image of the subject on the display unit 35 (step S1). Here, the through image is an image in which an image currently captured by the CCD 31 is displayed as it is. For example, an object imaged by the CCD 31 is stored in the DRAM 39, and the stored video is displayed on the display screen. It is realized by doing. Therefore, when the object imaged by the CCD 31 changes, the image stored in the DRAM 39 is updated, and the image displayed on the display screen changes from time to time. In the through display state, the video stored in the DRAM 39 is displayed as it is, but since the video data is not recorded on the memory card 42, the video data is not saved.

  Next, when it is determined that the user has pressed the shutter button halfway with the through image displayed (step S2), AF (Auto Focus) processing is started (step S3). When the AF process is completed, a through image display is performed (step S4). This through image display is performed until the shutter button is fully pressed (branch to NO in step S5).

  When the shutter button is fully pressed from the half-pressed state (step S4), still image shooting processing is started (step S5). Then, after taking a still image, a still image file is generated using the still image data developed on the buffer memory (DRAM 39), compressed and stored in the memory card 42 (step S6). When this still image file is compressed, an image quality priority compression mode described later is executed, and the still image file is compressed using a good quantization table. Then, it returns to step S1. By repeating this operation, still image files are sequentially recorded on the memory card 42, and the plurality of still image files are recorded on the memory card 42.

"Processing procedure by moving image shooting mode"
FIG. 8 is a flowchart showing the processing procedure of the DSP / CPU 32 when the user sets the moving image shooting mode.
When the user operates the key input unit 36 to set the moving image shooting mode, it is possible to take a scene or scenery that the user wants to take with the digital camera 30 and record the taken image on a memory card or the like. .

  That is, when the moving image shooting mode is set, the DSP / CPU 32 starts shooting by the CCD 31 and displays a through image of the subject on the display unit 35 (step S10). The through image is an image obtained by directly displaying an image currently captured by the CCD 31 as described above.

  Next, when it is determined that the recording start button is pressed by the user while the through image is displayed (step S11), a moving image (moving image frame) is captured at a predetermined frame rate (fixed period of 1/30 seconds). Then, the moving image shooting process to be stored is started (step S12). Through display continues until the recording start button is pressed.

  When the recording end button is pressed (step S13) or when a predetermined time (for example, 30 seconds) elapses after the moving image shooting process starts (step S14), the moving image shooting process is stopped (step S15). A moving image file is generated using the moving image data developed on the buffer memory (DRAM 39) until the time is compressed and stored in the memory card 42 (step S16). When the moving image file is compressed, an image quality priority compression mode, which will be described later, is executed, and the moving image file is compressed using a good quantization table. In particular, in the compression of moving images, the compression rate tends to be higher than that in the compression of still images. Therefore, in the processing in step S29 (see FIG. 10) in the image quality priority compression mode, which will be described later, the threshold value is still image shooting. A quantization table that is adjusted to a larger value than the time and has a higher compression rate than that during still image shooting is used.

  After performing the process of step 16, it returns to step S10. By repeating this operation, moving image files are sequentially recorded on the memory card 42, and the plurality of moving image files are recorded on the memory card 42. If the recording end button is not pressed in step S4 and 30 seconds have not elapsed since the moving image shooting process was started (NO in step S14), the process returns to step S13 until the recording end button is pressed or 30 seconds elapse. The moving image shooting process is continued.

`` Processing procedure with image quality priority compression mode ''
Next, FIG. 9 and FIG. 10 are flowcharts showing processing procedures in the image quality priority compression mode in the DSP / CPU 32.
First, the original image data is divided into a plurality of blocks, and each is subjected to discrete cosine transform (step S20). Next, a block containing a lot of high-frequency components is searched for from the transform coefficient obtained by the discrete cosine transform, and it is set as an image quality evaluation value calculation block BKh (step S21). Next, one of a predetermined number (for example, four) of quantization tables designated in advance is selected (step S22). Then, using the selected quantization table, the image data of the image quality evaluation value calculation block BKh and the neighboring blocks BKn1 to BKn4 adjacent to the block BKh are compressed, and each compressed image data is further expanded (step S23). .

  Then, the pixel value at the peripheral edge of the decompressed image quality evaluation value calculation block BKh and the image quality evaluation value calculation block BKh of each of the four neighboring blocks BKn1 to BKn4 adjacent to the image quality evaluation value calculation block BKh The sum (Σ) of the difference from the pixel value at the peripheral edge on the side in contact is obtained (step S24). Then, the obtained sum (Σ) is used as the image quality evaluation value of the currently selected quantization table (step S25). After calculating the image quality evaluation values of the currently selected quantization table, it is determined whether the image quality evaluation values have been calculated for all of a predetermined number (for example, four) of quantization tables specified in advance (step S26). . In this determination, if the calculation of the image quality evaluation value has not been completed for all the quantization tables, the next quantization table is selected (step S27). Then, the process returns to step S23, and the image quality evaluation value is obtained by compressing / decompressing the image data of the image quality evaluation value calculation block BKh and the neighboring blocks BKn1 to BKn4 adjacent thereto using the next quantization table. On the other hand, if it is determined in step S26 that the image quality evaluation values have been calculated for all the quantization tables, the processing of the flowchart shown in FIG. 10 is continued.

  After the calculation of the image quality evaluation values for all the quantization tables, the quantization is performed using an approximation method such as a cubic spline function based on the image quality evaluation values for the image quality evaluation value calculation block BKh for each quantization table. An approximate curve of a graph representing a change in image quality evaluation value when the table number is changed is calculated (step S28). Then, the calculated derivative value of the approximate curve is compared with a predetermined threshold value, and the quantization table whose slope is equal to or greater than the predetermined threshold value has the largest quantization table number, that is, the one with the lowest compression rate. A value larger by “1” than the quantization table number is detected and set as a quantization table number to be selected by the quantization table selection unit 21 (step S29). Then, the original image data is compressed and stored using the quantization table having the calculated quantization table number (step S30).

  As described above, the digital camera 30 according to the present embodiment has the same image compression function as that of the image compression apparatus according to the first embodiment. It is possible to compress an image in which image quality degradation hardly occurs even when the size is reduced, with a minimum file size without causing significant image quality degradation. As a result, more images can be recorded on the memory card 42 than before, and the memory card 42 can be used effectively.

  Further, an image quality evaluation value is obtained from the image quality evaluation value calculation block BKh for a block having a high frequency component among the original image data divided into a plurality of blocks, and a good quantization table is selected using the obtained image quality evaluation value. Therefore, the image compression processing time can be shortened.

  The present invention can be applied to an image photographing apparatus such as a digital camera or a video camera. In addition, a mobile phone having a photographing function, a clock, a PDA, or a personal computer having no photographing function may be used. In short, any device having a photographing function or a device capable of recording image data may be used. anything is fine.

1 is a functional block diagram of an image compression apparatus according to Embodiment 1 of the present invention. It is a figure for demonstrating the quantization table used for the image compression apparatus which concerns on Embodiment 1 of this invention. It is a figure for demonstrating the relationship between the quantization table used for the image compression apparatus which concerns on Embodiment 1 of this invention, and file size. It is a figure for demonstrating calculation of the image quality evaluation value in the image compression apparatus which concerns on Embodiment 1 of this invention. It is a figure which shows the approximated curve of the image quality evaluation value calculated with the image compression apparatus which concerns on Embodiment 1 of this invention. It is a block diagram which shows schematic structure of the digital camera which concerns on Embodiment 2 of this invention. It is a flowchart for demonstrating the process sequence by the still image shooting mode in the digital camera which concerns on Embodiment 2 of this invention. It is a flowchart for demonstrating the process sequence by the moving image imaging | photography mode in the digital camera which concerns on Embodiment 2 of this invention. It is a flowchart for demonstrating the process sequence by the image quality priority compression mode in the digital camera which concerns on Embodiment 2 of this invention. It is a flowchart for demonstrating the process sequence by the image quality priority compression mode in the digital camera which concerns on Embodiment 2 of this invention.

Explanation of symbols

DESCRIPTION OF SYMBOLS 9 Image data acquisition part 10 Block division part 11 DCT calculating part 12 Image quality evaluation block detection part 13 Quantization table part 14 Quantization process part 15 Entropy encoding part 16 Block input part 17 Entropy reverse encoding part 18 Inverse quantization part 19 Inverse DCT operation unit 20 Quantization table selection unit 21 Image quality evaluation value calculation unit 22 Threshold adjustment unit 23 Data output unit 30 Digital camera 32 DSP / CPU
42 Memory card

Claims (7)

Divide the original image data into multiple blocks,
Each of the plurality of divided blocks is subjected to discrete cosine transform to obtain a transform coefficient,
A block having a high frequency component is detected as a block for calculating an image quality evaluation value from the conversion coefficients of the plurality of blocks obtained thereby,
Using the plurality of quantization tables having different values of one or more elements in the quantization table, the detected image quality evaluation value calculation block and the surrounding blocks adjacent to the image quality evaluation value calculation block are compressed. ,
The compressed image quality evaluation value calculation block and surrounding blocks adjacent to the image quality evaluation value calculation block are expanded using the quantization table used in the compression,
The pixel value of the peripheral portion of the image quality evaluation value calculation block after expansion and the peripheral pixel value of the peripheral block adjacent to the image quality evaluation value calculation block on the side in contact with the image quality evaluation value calculation block Using the difference between and the image quality evaluation value for each quantization table,
Obtain an approximate curve that approximates the relationship between each quantization table and the image quality evaluation value,
Based on the approximate curve, determine a quantization table number corresponding to the quantization table used for compression of the original image,
An image compression method comprising compressing the original image data using a quantization table having the determined quantization table number. As a method of determining the quantization table number corresponding to the quantization table used for compression of the original image,
The differential value of the approximation curve corresponding to each quantization table is compared with a predetermined threshold value, among the quantization tables wherein the differential value is equal to or greater than a predetermined threshold value, decreases the value of each element of the quantization table is most 2. The image compression method according to claim 1, wherein a quantization table number is determined as a quantization table number corresponding to a quantization table used for compressing the original image. A block division process for dividing the original image data into a plurality of blocks;
DCT arithmetic processing for performing discrete cosine transform on each of the plurality of divided blocks;
An image quality evaluation value calculation block detection process for detecting a block having a high frequency component as a block for calculating an image quality evaluation value from the conversion coefficients of the plurality of blocks obtained by the discrete cosine transform;
Using a plurality of quantization tables having different values of one or more elements in the quantization table, each of the detected image quality evaluation value calculation block and each neighboring block adjacent to the image quality evaluation value calculation block A compression process for compressing each quantization table;
Decompression processing for decompressing each compressed image quality evaluation value calculation block and the image quality evaluation value calculation block using the same quantization table as in the compression,
The side in contact with the image quality evaluation value calculation block of the pixel value at the peripheral edge of the image quality evaluation value calculation block after expansion for each quantization table and the surrounding blocks adjacent to the image quality evaluation value calculation block An image quality evaluation value calculation process for calculating an image quality evaluation value for each quantization table using a difference from the pixel value of the peripheral edge of
An approximate curve calculation process for calculating an approximate curve that approximates the relationship between each quantization table and the image quality evaluation value;
Based on the calculated approximate curve, a quantization table number determination process for determining a quantization table number corresponding to a quantization table used for compression of the original image;
Original image data compression processing for compressing the original image data using a quantization table having the determined quantization table number;
Including
An image compression program for compressing image data of a subject photographed by an image photographing device by a computer. As a method of determining the quantization table number corresponding to the quantization table used for compression of the original image,
The differential value of the approximate curve corresponding to each quantization table is compared with a predetermined threshold, and among the quantization tables in which the differential value is equal to or greater than the predetermined threshold, the value of each element of the quantization table is the smallest. 4. The image compression program according to claim 3, wherein a quantization table number is determined as a quantization table number corresponding to a quantization table used for compressing the original image. Block dividing means for dividing the original image data into a plurality of blocks;
DCT computing means for obtaining a transform coefficient by performing discrete cosine transform on each of the plurality of blocks divided by the block dividing means;
Image quality evaluation value calculation block detection means for detecting a block having a high frequency component from the conversion coefficient of each of the plurality of blocks obtained by the DCT calculation means, as an image quality evaluation value calculation block;
The image quality evaluation value calculation block detected by the image quality evaluation value calculation block detection means and the image quality evaluation value calculation using a plurality of quantization tables having different values of one or more elements in the quantization table Compressing the surrounding blocks adjacent to the image block, and further compressing the compressed image quality evaluation value calculating block and the surrounding blocks adjacent to the image quality evaluation value calculating block used in the compression Compression / expansion means for expanding using
The pixel value of the peripheral portion of the image quality evaluation value calculation block after expansion and the peripheral pixel value of the peripheral block adjacent to the image quality evaluation value calculation block on the side in contact with the image quality evaluation value calculation block Image quality evaluation value calculation means for obtaining an image quality evaluation value for each quantization table using the difference between
An approximate curve calculating means for obtaining an approximate curve that approximates the relationship between each quantization table and the image evaluation value;
A quantization table number determining unit that determines a quantization table number corresponding to a quantization table used for compression of the original image based on the approximate curve obtained by the approximate curve calculating unit;
Original image data compression means for compressing the original image data using the quantization table of the quantization table number determined by the quantization table number determination means;
An image compression apparatus comprising: The quantization table number determining means includes
Comparing the derivative of the approximate curve corresponding to each quantization table with a predetermined threshold;
Wherein among the quantization table differential value is equal to or greater than a predetermined threshold value, a quantization table corresponding to the smallest quantization table number is the value of each element of the quantization table in quantization table used for compression of the original image 6. The image compression apparatus according to claim 5 , wherein the image compression apparatus is determined as a number. A digital camera comprising the image compression apparatus according to claim 5 .

Images