JP4939460B2 - Image processing device - Google Patents

Description

  The present invention relates to an image processing apparatus for generating compressed data of image data.

  Many image input devices such as scanners and digital cameras take captured images as image data, and compress and encode the image data using a JPEG (Joint Photographic Experts Group) compression method. In general, in the JPEG compression method, as shown in Patent Document 1, input image data is converted from RGB format to YUV format, and pixel data of n × n (n is a natural number, mostly n = 8 is used). Then, a discrete cosine transform (hereinafter referred to as “DCT (Discrete Cosine Transform)”), which is a kind of orthogonal transform, is performed.

  Data subjected to the DCT transformation is quantized using a quantization table. This quantization is performed by dividing the DCT transformed data by the quantization constant of the quantization table. Here, the magnitude of the quantization constant of the quantization table determines the quantization step. That is, if the quantization constant is increased, the quantization step is increased and the compression rate is increased. On the other hand, the image quality of the compressed image data is deteriorated. On the contrary, if the quantization constant is reduced, the quantization step is reduced and the compression rate is lowered. However, the image quality deterioration of the image data after compression can be suppressed.

Then, the quantized data is encoded to generate compressed data. For encoding, Huffman encoding, which is a kind of entropy encoding, is mainly used.
JP 2000-184330 A

  However, in the case of the general JPEG compression method as described above, in many cases, quantization is performed using a common quantization table for the Y component, U component, and V component. In other words, regardless of the ratio of the Y component, U component, and V component with respect to the image data, all components are compressed at the same compression rate. For example, if the ratio of the Y component to the image data is very high compared to the U component and the V component, but all the components are compressed at the same compression rate, the Y component is compressed with a low image quality. Become. This has been a cause of lowering the image quality of the entire image.

  The present invention has been made to solve the above problem, and aims to improve the compression ratio while maintaining the image quality after compression by changing the compression ratio for each component after YUV conversion. Is.

  The image processing apparatus according to claim 1 is a conversion unit that converts image data into a color system composed of a luminance component and a color difference component; a histogram generation unit that generates a histogram of the luminance component and the color difference component; From the created histogram, ratio calculation means for calculating the ratio of the luminance component and the ratio of the color difference component to the image data, orthogonal transformation means for orthogonal transformation of the transformed image data, and different quantization constants A storage means for storing a plurality of set quantization tables; a quantization table for the luminance component from the plurality of stored quantization tables according to a preset image quality level and the calculated ratio; Selection means for selecting the quantization table for each of the color difference components, and the luminance component of the orthogonally transformed image data as the luminance Quantization means for quantizing using a quantization table for quantization, and quantizing the chrominance component using the quantization table for the chrominance component, and encoding the quantized image data and outputting compressed data And encoding means.

  According to this configuration, it is possible to select a quantization table having a different quantization constant for each component according to a preset image quality level and the ratio of each component to image data. Therefore, data compression can be performed at a compression rate suitable for the ratio of each component.

  A second aspect of the present invention is the image processing apparatus according to the first aspect, wherein the selection unit is configured to detect a component having a low ratio to the image data among the luminance component and the color difference component. A quantization table having a larger quantization constant than a quantization table selected for a component having a high ratio to image data is selected.

  According to this configuration, it is possible to perform quantization with a low compression ratio for a component with a high ratio and with a high compression ratio for a component with a low ratio. That is, since a component with a high ratio is compressed at a low compression rate, deterioration in image quality after compression can be suppressed, and a component with a low ratio is compressed at a high compression rate, so that the compression rate can be improved. Therefore, as a whole, compression can be performed at a high compression rate while suppressing deterioration in image quality after compression.

  Invention of Claim 3 is invention of Claim 1 or 2, Comprising: The decoding means which decodes the output said compression data, and outputs decompression | restoration image data, The image quality of the said decompression | restoration image data is Comparing means for comparing whether or not the image quality level has been reached, and when the image quality of the restored image data has not reached the image quality level, the selection means sends a ratio of the luminance component and the color difference component to the image data. A quantization table having a smaller quantization constant than the quantization table for the higher component, and the quantization means quantizes the luminance component and the color difference component using the selected quantization table. Further, it is characterized by further comprising repetition control means for causing the encoding means to encode the quantized image data and output compressed data.

  According to this configuration, since the quantization table is switched and selected according to the ratio of each component and compression processing is performed until the image quality of the restored image data reaches a preset image quality level, compression for each component is performed. The rate can be adjusted efficiently.

  According to the present invention, it is possible to select a quantization table having different quantization constants according to a preset image quality level and a ratio of each component to image data, so that a compression rate suitable for the ratio of each component can be selected. Data compression can be performed. Furthermore, since a component with a high ratio is compressed at a low compression rate, image quality deterioration after compression can be suppressed, and a component with a low ratio is compressed at a high compression rate, so that the compression rate can be improved. Therefore, as a whole, compression can be performed at a high compression rate while suppressing deterioration in image quality after compression.

  An image processing apparatus according to an embodiment of the present invention will be described with reference to the drawings. FIG. 1 is a functional block diagram of an image processing apparatus 1 according to the present embodiment. The image processing apparatus 1 compresses input image data using a JPEG compression method, and includes a control unit 11, a data input / output unit 12, a YUV conversion unit 13, a histogram creation unit 14, and a DCT conversion unit 15. A storage unit 16, a quantization unit 17, an encoding unit 18, a decoding unit 19, and the like.

  The data input / output unit 12 receives image data input from the outside and outputs it to the control unit 11. Further, under the control of the control unit 11, compressed data and decoded image data are output to the outside. The YUV conversion unit 13 converts the input image data from the RGB format to the YUV format.

  The histogram creation unit 14 generates a histogram of Y component, U component, and V component as shown in FIGS. The DCT conversion unit 15 converts the sub-blocks divided into n × n pixel data into Y × n pixel data after the YUV conversion from the spatial domain to the frequency domain DCT coefficients using DCT conversion. The DCT conversion unit 15 repeatedly performs DCT conversion on the divided sub-blocks over one image (one frame).

  The storage unit 16 stores the quantization table group 2 and stores programs, data, and the like for realizing various functions of the image processing apparatus 1. The quantization table group 2 stores a plurality of quantization tables having different quantization constants. FIG. 2 is a diagram illustrating a data configuration of the quantization table group 2 stored in the storage unit 15. When the control unit 11 divides the YUV-converted image data into, for example, 8 × 8 sub-blocks, the quantization table has an 8 × 8 quantization constant that is the same number of matrices as the sub-blocks. In general, in order to increase the compression rate while maintaining the image quality, the quantization constant corresponding to the upper left side of the matrix is set small, and the quantization constant corresponding to the lower right side is set large.

  Each of the quantization tables 21, 22,..., N (N is an integer equal to or greater than 1; hereinafter referred to as “quantization table 20”) is a quantum when the quantization unit 17 quantizes the sub-block. Table numbers are stored in association with each other in descending order of the conversion step (that is, from the higher compression ratio). FIG. 3 shows an example of the relationship between the table number of the quantization table and the compression rate. For example, if a sub-block is quantized and encoded using the quantization table (quantization table 21) of table number 1, compressed data with a compression rate of 5% is generated, and the quantization table (quantum table of table number 2) When the sub-block is quantized and encoded using the conversion table 22), compressed data with a compression rate of 10% is generated.

  Returning to FIG. The quantization unit 17 quantizes the DCT-transformed sub-block using the quantization table selected by the quantization table selection unit 111. The encoding unit 18 encodes the sub-block quantized by the quantization unit 17 by entropy encoding using Huffman encoding, and outputs the encoded data to the control unit 11 as compressed data. Specifically, the coefficient sequence of the sub-block is zigzag scanned, and a known run-length encoding is performed on consecutive 0s, and variable-length encoding or Huffman encoding is performed.

  The decoding unit 19 decodes the compressed data using means opposite to the image data compression means described above, and generates decompressed compressed data. Note that a detailed description of decoding is omitted because a known technique is used.

  The control unit 11 is configured by a CPU (Central Processing Unit) or the like, and performs overall control of the image processing apparatus 1 by outputting instruction signals to each function unit, transferring data, and the like. The control unit 11 also functions as a ratio calculation unit 111, a quantization table selection unit 112, and an image quality comparison unit 113.

  Specifically, when the data input / output unit 12 inputs image data, the control unit 11 causes the YUV conversion unit 13 to YUV convert the image data, and causes the histogram creation unit 14 to generate a histogram of each component. Then, the ratio calculation unit 111 calculates the ratio of each component to the image data.

  An example of a method for calculating the ratio of each component by the ratio calculation unit 111 will be specifically described. Here, the ratio means the ratio of the information amount of each component to the information amount of image data for one frame (one image).

  4A shows an example of a Y component histogram, FIG. 4B shows an example of a U component histogram, and FIG. 4C shows an example of a V component histogram. The ratio calculation unit 111 calculates the total sum of frequencies in each Y value, each U value, and each V value using the histogram of each component, and calculates the total value of each sum. Then, the ratio calculation unit 111 calculates the ratio of the Y values using the formula of (the sum of the frequencies at each Y value / total value), and then (the total number / the total value of the frequencies at each U value), (each The ratio of the U value and the V value is calculated using a formula of (total number of frequencies in V value / total value).

  Needless to say, the sum of the frequency of each value is proportional to the area of each histogram (area of the filled area of the graph). Assuming that the area of the histogram of each YUV component shown in FIG. 4 is substantially the same, the total number of frequencies of each value is considered to be substantially the same. In this case, an equivalent ratio (for example, about 30%) is calculated for each component. On the other hand, as shown in FIG. 5, it is assumed that the area of the histogram of each UV component is substantially the same, but the area of the Y component is about twice the area of the histogram of each UV component. In this case, for example, a ratio of 50% for the Y component and 25% for each UV component is calculated.

  Then, in accordance with the ratio of each component and a preset image quality level (hereinafter referred to as “set image quality level”), the quantization table selection unit 112 selects each YUV from the quantization table group 2 stored in the storage unit 16. Select a quantization table to be used when quantizing components. Specifically, first, a quantization table is selected based only on the set image quality level. For example, when a quantization table with a compression rate of 30% is selected based on the set image quality level, if the ratio of each component is the same, the quantization table selection unit 112 converts the quantization table with a compression rate of 30% to each YUV. Select as a quantization table for components.

  On the other hand, when the ratio of 50% for the Y component and 25% for each UV component is calculated, the quantization table selection unit 112 uses the compression rate of the quantization table selected based on the set image quality level as a reference, for example, A quantization table with a compression rate of 40% is selected as the quantization table for the Y component, and a quantization table with a compression rate of 20% is selected as the quantization table for each UV component. That is, since the ratio of each UV component to the ratio of the Y component is 1/2, the quantization table selection unit 112 doubles the compression ratio of the quantization table for the Y component as the quantization table for the UV component. Select one with a compression ratio of.

  In this way, by selecting a quantization table with a different compression ratio according to the ratio of each component, a component with a higher ratio is quantized with a lower compression ratio than a component with a lower ratio. Image quality degradation after compression of high components can be suppressed, and components with lower ratios can be compressed with a high compression rate, so that the quality of the restored image can be maintained while increasing the compression rate as a whole.

  Returning to FIG. After the quantization table for each component is selected by the quantization table selection unit 112, the control unit 11 divides the YUV converted data into sub-blocks made up of n × n pixel data, and performs DCT conversion for each sub-block. The unit 14 performs DCT conversion. Then, the quantizing unit 17 quantizes the DCT-transformed sub-block using the quantization table, and causes the coding unit 18 to encode the quantized sub-block to generate compressed data.

  Then, the control unit 11 causes the decoding unit 19 to decode the compressed data and output the restored image data, and the image quality comparing unit 113 compares the image quality of the uncompressed image data and the restored image data. Specifically, the image quality comparison unit 113 performs image quality evaluation by calculating a PSNR (Peak Signal Noise Ratio) value, for example. The PSNR value is a value obtained by averaging the difference relative ratio of each pixel over all pixels between the input original image and the restored image obtained by decoding the compressed data of the original image. Is used.

The image quality comparison unit 113 uses the PSNR value to determine whether the restored image data has reached the set image quality level.

  In accordance with this determination result, the quantization table selection unit 111 switches and selects the quantization table for each component, and the control unit 11 switches to the quantization unit 17 to perform quantization with the selected quantization table, The encoding unit 18 encodes the compressed data. The controller 11 repeats this control until the image quality of the restored image data reaches the set image quality level. Details of this repetitive control will be described later.

  6 and 7 are flowcharts showing the flow of compression processing by the image processing apparatus 1. When the data input / output unit 12 inputs image data, the control unit 11 performs YUV conversion on the image data input to the YUV conversion unit 13 (step S11), and causes the histogram creation unit 14 to create a histogram of each component (step S12). . Then, the ratio calculation unit 111 calculates the ratio of each component in the image data (step S13). Next, the control unit 11 divides the signal-converted image data into sub-blocks made up of n × n pixel data, and causes the DCT conversion unit 14 to DCT-convert each sub-block (step S14). Here, the control unit 11 stores each DCT-converted sub-block in a working memory (not shown) included in the control unit 11.

  Next, the quantization table selection unit 112 selects a quantization table for each component from the quantization table group 2 according to the set image quality level and the ratio of each component (step S15). Here, the control unit 11 assigns the table number of the quantization table for the Y component selected by the quantization table selection unit 112 to the variable Iy, and assigns the table number of the quantization table for the U component to the variable Iu. Then, the table number of the quantization table for the V component is substituted into the variable Iv. Variables Iy, Iu, and Iv are variables set in the working memory of the control unit 11. In step S15, the quantization table selection unit 112 may select the quantization table for each component so that the image quality level of the restored image data after being compressed and decoded is slightly higher than the set image quality level. desirable.

  Then, the control unit 11 quantizes each sub-block (step S16), causes the encoding unit 18 to encode each quantized sub-block and outputs compressed data (step S17). Further, the control unit 11 causes the decoding unit 19 to decode the compressed data and output restored image data (step S18).

  Next, the image quality comparison unit 113 compares the image quality of the input image data and the restored image data, and determines whether or not the image quality of the restored image data has reached the set image level (step S19). When the image quality of the restored image data is at the set image level (step S19; set image quality level), the control unit 11 outputs the compressed data to the outside via the data input / output unit 12, and ends the control.

  The image quality of the restored image data is above the set image level (step S19; above the set image quality level), the variable f is 0 (step S20; 0), and any of the variables Iy, Iu, and Iv is 1. If it is not (that is, the first table number of the quantization table 20) (step S21; NO), the control unit 11 subtracts 1 from the variables Iy, Iu, and Iv (step S22), and shifts the processing to step S16. That is, the quantization table selection unit 112 switches and selects a quantization table whose compression ratio is one step higher for each component, and the control unit 11 switches to the quantization unit 17 and uses the selected quantization table. In step 14, each DCT transformed sub-block stored in the working memory is quantized. Note that the variable f is a variable set in the working memory of the control unit 1, and the contents of the numerical value substituted for the variable f will be described later.

  When any of the variables Iy, Iu, and Iv is 1 (step S21; YES), the quantization table group 2 is a quantization table having a larger quantization constant (ie, can be quantized with a higher compression rate). Since the (quantization table) is not stored, compression processing at a higher compression rate cannot be performed. Therefore, the control unit 11 outputs the compressed data output in step S17 to the outside via the data input / output unit 12, and ends the control. Note that a quantization table having a higher compression rate may be generated by multiplying each quantization constant of the quantization table of table number 1 by a constant.

  As shown in steps S16 to S22, while the image quality of the restored image data is above the set image quality level, the quantization table selection unit 112 converts the quantization table for each component into a quantization table with a high compression rate by one step. The control unit 11 performs control to repeatedly perform quantization, encoding, and decoding using the selected quantization table.

  When the quantization table is switched and selected in this way, the timing at which the image quality of the restored image data falls below the set image quality level occurs. That is, in step S19, it is determined that the image quality of the restored image data is lower than the set image level, and any of the variables Iy, Iu, and Iv is not N (that is, the last table number of the quantization table 20). (Step S23; NO) When the variable f is 0 (Step S24; 0), the control unit 11 adds 1 to the table number of the component having the highest ratio calculated in Step S13, and further sets the variable f to 1. Is substituted (step S25), and the process proceeds to step S16. That is, the quantization table selection unit 112 selects and selects the component with the highest ratio by switching to a quantization table with a lower compression ratio.

  Here, the variable f indicates which component quantization table is individually switched. When f = 0, the quantization table set in step S15 is selected. When f = 1, it indicates that the quantization table of the component with the highest ratio has been switched, and f = 2. Indicates that the quantization table of the component with the second highest ratio has been switched.

  When the variable f is 1 (step S24; 1), the control unit 11 adds 1 to the table number of the component having the second highest ratio, and further substitutes 2 for the variable f (step S26). Transfer processing to.

  When the variable f is 2 (step S24; 2), since the quantization table having different compression ratios for the two components has already been switched and selected, the control unit 11 passes through the data input / output unit 12. To output the compressed data to the outside and finish the control.

  In step S23, when the control unit 11 determines that any of the variables Iy, Iu, and Iv is N (step S23; YES), the quantization table group 2 is quantized having a smaller quantization constant. Since a table (that is, a quantization table that can be quantized at a lower compression rate) is not stored, compression processing at a lower compression rate cannot be performed. Therefore, the control unit 11 outputs the compressed data to the outside via the data input / output unit 12 and ends the control. Note that a quantization table having a lower compression rate may be generated by dividing each quantization constant of the quantization table having the table number N by a constant.

  Further, in step S19, the image quality comparison unit 113 determines that the image quality of the restored image data is above the set image quality level (step S19; above the set level), and the control unit 11 determines that the variable f is 1 or 2. When the determination is made (step S20; 1 or 2), the control unit 11 outputs the compressed data to the outside via the data input / output unit 12, and ends the control.

  As described above, by calculating the ratio of each component of YUV to image data and switching and selecting the quantization table for each component according to this ratio and the set image quality level, the component with a high ratio is obtained. Quantization can be performed with a low compression rate and with a high compression rate for components with a low ratio. That is, image data can be compressed at a high compression rate while suppressing deterioration in image quality after compression.

The block diagram which shows the internal structure of an image processing apparatus. The figure which shows the data structure of a memory | storage part. An example of the histogram of each component. An example of the histogram of each component. The figure which shows an example of the relationship of the compression rate with the table number of a quantization table. The flowchart which shows the flow of a compression process. FIG. 7 is a continuation flowchart of FIG. 6.

Explanation of symbols

1 Image Processing Device 11 Control Unit (Repetitive Control Unit)
111 Ratio calculation part (ratio calculation means)
112 Quantization table selection unit (selection means)
113 Image quality comparison unit (comparison means)
12 Data input / output unit 13 YUV conversion unit (conversion means)
14 Histogram creation part (histogram creation means)
15 DCT transform unit (orthogonal transform means)
16 Storage unit (storage means)
2 Quantization table group 20 Quantization table 17 Quantization unit (quantization means)
18 Encoding unit (encoding means)
19 Decoding unit (decoding means)

Claims (3)

Conversion means for converting image data into a color system comprising a luminance component and a color difference component;
Histogram creation means for creating a histogram of the luminance component and the color difference component;
Ratio calculation means for calculating a ratio of the luminance component and a ratio of the color difference component with respect to the image data from the created histogram,
Orthogonal transformation means for orthogonal transformation of the transformed image data;
Storage means for storing a plurality of quantization tables set with different quantization constants;
Selection means for selecting the luminance component quantization table and the color difference component quantization table from the plurality of stored quantization tables, respectively, in accordance with a preset image quality level and the calculated ratio. When,
Quantization means for quantizing the luminance component of the orthogonally transformed image data using a quantization table for the luminance component and quantizing the color difference component using the quantization table for the color difference component; ,
Encoding means for encoding the quantized image data and outputting compressed data;
An image processing apparatus.   The selection means has a larger quantization constant than a quantization table selected for a component having a low ratio to the image data among the luminance component and the color difference component for a component having a high ratio to the image data. The image processing apparatus according to claim 1, wherein a quantization table is selected. Decoding means for decoding the output compressed data and outputting restored image data;
Comparison means for comparing whether or not the image quality of the restored image data has reached the image quality level;
When the image quality of the restored image data does not reach the image quality level, a quantization table for a component having a high ratio of the luminance component and the color difference component to the image data is quantized with a small quantization constant. A table for switching and selecting, the luminance means and the color difference component are quantized using the quantization table selected by switching to the quantizing means, and the quantizing means encodes the quantized image data Repetitive control means for outputting the compressed data,
The image processing apparatus according to claim 1, further comprising: