JP2006165609A - Image processing apparatus, image processing method, and image processing program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a technology of establishing the compatibility between reduction in block noise and high-speed processing in reducing a compressed image obtained by orthogonal transformation and reproducing the reduced image. <P>SOLUTION: The image processing apparatus for generating the reduced image data from the compressed image obtained by applying orthogonal transformation to image data 1 formed from a space region into a frequency region includes: an IDCT processing section 3 that applies inverse orthogonal transformation to M row × M column data (M is a natural number) of orthogonal transformation coefficients to produce data of M row × M column pixels; a DC component extract section 4 for acquiring a DC component in the orthogonal transformation coefficients to produce data of 1×1 pixels; a smoothing processing section 5 that combines the data of the M row × M column pixels generated by the IDCT processing section 3 with the DC component extracted by the DC component extract section 4 and smoothes the data of the M row × M column pixels as required to produce image data; and an optional size reduction section 6 for generating the reduced image data with an optional size from the image data generated by the smoothing processing section 5. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to an image processing apparatus, an image processing method, and an image processing program capable of removing block noise when generating a reduced image from a compressed image such as DCT (Discrete Cosine Transform).

  Conventionally, in Patent Document 1 (Japanese Patent Laid-Open No. 2003-234907), a low-frequency component and a high-frequency component of a DCT coefficient are separated by JPEG (Joint Photographic Experts Group) expansion processing, and the low-frequency component is separated from a plurality of adjacent blocks. Block noise is reduced by substituting values obtained by smoothing. Moreover, in patent document 2 (Unexamined-Japanese-Patent No. 2001-189937), when processing an image by a two-dimensional block unit, the information of the weighting coefficient for smoothing is added to each block, and according to the value of a coefficient Block noise is removed by changing the smoothing degree.

Japanese Patent Laid-Open No. 2003-234907 JP 2001-189937 A

  However, the method described in Patent Document 1 has a problem in that not only block noise but also edge information is removed because a value obtained by smoothing a low frequency component of an adjacent block is used. On the other hand, there is a problem that noise increases in the block close to the edge. Further, in the method described in Patent Document 2, it is necessary to add weighting coefficient information for performing smoothing for each block in advance, and the processing becomes complicated. Is not suitable.

  The present invention has been made to solve such problems. That is, the present invention is an image processing apparatus for generating reduced image data from a compressed image obtained by orthogonally transforming image data from a spatial domain to a frequency domain, and M rows × M columns data (M is a natural number) of orthogonal transform coefficients. An inverse transform processing unit that generates M × M pixel data by inverse orthogonal transformation, a DC component extraction unit that obtains a DC component from the orthogonal transform coefficient and generates 1 × 1 pixel data, and inverse transform Smoothing that combines the M row × M column pixel data generated by the processing unit and the DC component extracted by the DC component extraction, and smoothes the M row × M column pixel data as necessary to generate image data. And a reduction unit that generates reduced image data of an arbitrary size from the image data generated by the smoothing unit.

  The present invention is also an image processing method for generating reduced image data from a compressed image obtained by orthogonally transforming image data from a spatial domain to a frequency domain, and M rows × M columns data (M is a natural number) of orthogonal transform coefficients. Are obtained by performing inverse orthogonal transform to generate M row × M column pixel data, obtaining a DC component from the orthogonal transform coefficient and generating 1 × 1 pixel data, and M obtained by inverse orthogonal transform. A process of combining row x M column pixel data and a direct current component and smoothing M row x M column pixel data as necessary to generate image data, and a reduced image of any size from the generated image data Generating data.

  The present invention is also an image processing program for generating reduced image data from a compressed image obtained by orthogonally transforming image data from a spatial domain to a frequency domain, and M rows × M columns data (M is a natural number) of orthogonal transform coefficients. Are obtained by performing inverse orthogonal transformation to generate M row × M column pixel data, obtaining a DC component from the orthogonal transformation coefficient and generating 1 × 1 pixel data, and M obtained by inverse orthogonal transformation. A step of combining row x M column pixel data and a DC component and smoothing M row x M column pixel data as necessary to generate image data, and a reduced image of any size from the generated image data And a step of generating data.

  In the present invention, M × M data having a certain size of orthogonal transform coefficient is inversely orthogonal transformed to obtain M × M pixel data, and 1 × 1 pixel data based on a DC component is obtained. In order to generate and combine these, and smooth the data of M rows × M columns pixels as necessary, by using data of M rows × M columns pixels compared to the case of generating a reduced image with only a DC component. The reproducibility of the edge component is excellent, and the calculation time can be shortened by using the DC component.

  Therefore, according to the present invention, in reducing and reproducing an image compressed by orthogonal transform, a process for generating a reduced image of 1/2, 1/4, 1/8, etc. from the low frequency component of the orthogonal transform coefficient Both the high-speed image reduction processing and the block noise reduction by switching the block noise removal method based on the quantization table value and the reduction rate. Can be realized.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings. FIG. 1 is a block diagram illustrating an image processing apparatus according to this embodiment. The image processing apparatus according to the present embodiment generates reduced image data from a compressed image by, for example, DCT (Discrete Cosine Transform), which orthogonally transforms image data from a spatial domain to a frequency domain. 2, an IDCT (Inverse DCT) processing unit 3, a DC (direct current) component extraction unit 4, a smoothing processing unit 5, and an arbitrary size reduction unit 6.

  Input data 1 includes DCT-compressed image data such as a JPEG (Joint Photographic Experts Group) image and size information after reduction, and is passed to the reduction processing control unit 2. The reduction process control unit 2 obtains a reduction rate from the input image width and height information and the reduced image size information, and selects each of the IDCT process, the smoothing process, and the arbitrary size reduction process.

  First, the vertical and horizontal reduction ratios are obtained from the width and height of the input image and the size after reduction. Next, the larger one of the vertical and horizontal reduction ratios applies to any of (1) greater than 1/2, (2) 1/2 or less, (3) 1/4 or less, and (4) 1/8 or less. In case of (1), 8 × 8 IDCT processing unit, in case of (2) 4 × 4 IDCT processing unit, in case of (3) 2 × 2 IDCT processing unit, in case of (4) DC component extraction unit Select.

  As shown in FIG. 2, the IDCT processing unit 3 is an 8 × 8 IDCT processing unit 32 that performs 8 × 8 size IDCT processing, a 4 × 4 IDCT processing unit 33, a 2 × 2 IDCT processing unit 34, and these three types of IDCT processing. The process selection part 31 to select is provided.

When the 8 × 8 size DCT coefficient is subjected to IDCT processing, image data of 8 × 8 pixels is obtained as shown in FIG. The IDCT processing of the Y component is expressed by Equation 1. Here, the DCT coefficient of the u and v coordinates is F _Y (u, v).

  Similarly, when DCT coefficients of 4 × 4 size and 2 × 2 size are subjected to IDCT processing, image data of 4 × 4 pixels and 2 × 2 pixels are obtained as shown in FIGS. 3B and 3C, respectively.

  Next, the smoothing processing unit 5 combines the image data of a predetermined size and the image data based on the DC component, and performs the smoothing process according to the value of the quantization table. First, the average value of the quantization table is calculated, and if the value is equal to or greater than the threshold value, it is determined that many low-frequency components are included and the edge information is small, and smoothing processing is performed. On the contrary, when the average value of the quantization table is less than the threshold value, it is determined that a lot of edge information is included, and the smoothing process is not performed. Further, instead of calculating the average value of the quantization table, the quantization table value of the DC component may be used.

  In the smoothing process, first, DC components of adjacent blocks on the left and right are extracted and compared with the DC component values of the block of interest. When both the difference from the left block and the difference from the right block are equal to or less than the threshold value, it is determined that the block does not include edge information, and smoothing is performed. Conversely, if the difference is greater than or equal to the threshold, it is determined that the block includes edge information, and smoothing is not performed.

  When smoothing processing is performed, a difference d between a value obtained by averaging the DC components of the left and right blocks and a DC component value of the target block is obtained, and a value obtained by multiplying the smoothing coefficient s (0 ≦ s ≦ 1) by the smoothing coefficient Add to DC component. These processes are performed for all blocks. However, the left and right end blocks are not smoothed.

  When smoothing is not performed, image data of a predetermined size and image data based on DC components are arranged in a checkered pattern. For example, when 1/4 reduction is performed, 2 × 2 pixel data obtained by IDCT processing and 2 × 2 pixel data generated by the DC component are alternately arranged for each block. Thereby, the edge component is reproduced by 2 × 2 pixel data by the IDCT process, and the processing speed can be increased by using the 2 × 2 pixel data generated by the DC component.

  The DC component extraction unit 4 extracts the DC component of the DCT coefficient and performs IDCT processing. When the reduction ratio is 1/8 or less, 1 × 1 pixel data subjected to IDCT processing is output for 8 × 8 DCT coefficient data. The IDCT processing of only the DC component is expressed by Equation 2.

  Next, smoothing processing is performed in the same manner as the IDCT processing unit. However, the smoothing process is always performed here regardless of the value of the quantization table. The content of the process is the same as the smoothing process in the IDCT processing unit.

  Further, the smoothing processing unit 5 smoothes the pixel values at the boundary portion with the adjacent block according to the image reduction rate and the average value of the quantization table. In the case where the smoothing process is performed by the IDCT processing unit 3 or the DC component extraction unit 4, the smoothing is not performed here. Even when the reduction ratio is 1/8 or less and the average value of the quantization table is equal to or greater than the threshold, the block boundary portion is not smoothed.

  The arbitrary size reduction unit 6 reduces the image subjected to IDCT conversion by the IDCT processing unit or the DC component extraction unit to an arbitrary size. As an interpolation method for reduction, linear interpolation can be considered, but a projection method or the like may be used. By performing these processes, an output image reduced to a desired size is obtained.

  The flow of the above processing is shown in a flowchart in FIG. First, the compression rate is calculated from the reduced image size by the reduction processing control unit 2 (step S1). Next, the size of the IDCT process is selected according to the calculated compression rate (step S2), and the selected IDCT process is performed (step S3).

  Next, the value of the DC component in the block of interest and the value of the DC component in the adjacent block are compared to determine whether or not the difference is greater than or equal to a threshold value (step S4). Is smoothed by the smoothing processing unit 5 (step S5).

  On the other hand, if it is greater than or equal to the threshold value, the process proceeds to step S6 to determine whether or not the quantization table value is greater than or equal to the threshold value. If it is less than the threshold value here, it is determined that the edge component is not included, and smoothing is performed by the smoothing processing unit 5 (step S7). On the other hand, if the value is equal to or greater than the threshold value, the smoothing process is not performed because the edge component is included. Thereafter, the image data is reduced to an arbitrary size (step S8), and desired reduced image data is output. The above processing is performed on all the blocks to generate a reduced image in which portions other than the edge portions are smoothed.

  In the image processing method according to the flowchart, each step can be realized by program processing. That is, by realizing this image processing method by program processing, the image processing method of this embodiment is incorporated as a program as a function of graphics software executed by a personal computer or the like, or the image processing method of this embodiment Alternatively, it is possible to store and distribute software in which this is incorporated in a medium such as a CD-ROM or DVD-ROM, or to distribute it via a network.

In the above-described embodiment, an example in which a block of a compressed image after orthogonal transform is an 8 × 8 size is mainly used as an example, but the present invention is not limited to this, and 2 ⁿ × 2 ⁿ Any block of size (n is a natural number) can be applied. In this case, the IDCT processing section may be prepared with processing sections of 2 ^n-1 × 2 ^n-1 size to 2 × 2 size.

It is a block diagram explaining the image processing apparatus which concerns on this embodiment. It is a block diagram explaining the internal structure of an IDCT process part. It is a schematic diagram explaining the IDCT process of each size. It is a flowchart explaining the image processing method which concerns on this embodiment.

Explanation of symbols

  DESCRIPTION OF SYMBOLS 1 ... Input data, 2 ... Reduction process control part, 3 ... IDCT process part, 4 ... DC component extraction part, 5 ... Smoothing process part, 6 ... Arbitrary size reduction part, 7 ... Output data, 31 ... Process selection part, 32... 8 × 8 IDCT processing section, 33... 4 × 4 IDCT processing section, 34.

Claims (15)

In an image processing apparatus that generates reduced image data from a compressed image obtained by orthogonally transforming image data from a spatial domain to a frequency domain,
An inverse transform processing unit that performs inverse orthogonal transform on M row × M column data (M is a natural number) of orthogonal transform coefficients to generate M row × M column pixel data;
A DC component extraction unit that acquires a DC component of the orthogonal transform coefficient and generates 1 × 1 pixel data;
The M row × M column pixel data generated by the inverse conversion processing unit is combined with the DC component extracted by the DC component extraction, and the M row × M column pixel data is smoothed as necessary. A smoothing processing unit for generating image data;
An image processing apparatus comprising: a reduction unit that generates reduced image data of an arbitrary size from the image data generated by the smoothing processing unit. The image processing apparatus according to claim 1, further comprising a reduction processing control unit that selects the value of the M rows × M columns in the inverse conversion processing unit according to a reduction rate of the image data. The image processing apparatus according to claim 1, wherein the inverse transform processing unit performs inverse transform of a plurality of sizes of the M rows × M columns. The image processing apparatus according to claim 1, wherein the smoothing processing unit extracts an edge portion in a block of the image data, and smoothes only a block determined to have no edge portion. The image processing apparatus according to claim 1, wherein the smoothing processing unit selects whether to smooth the image data based on a reduction ratio and a quantization table value. In an image processing method for generating reduced image data from a compressed image obtained by orthogonally transforming image data from a spatial domain to a frequency domain,
M-row × M-column data (M is a natural number) of orthogonal transform coefficients and inverse orthogonal transform to generate M-row × M-column pixel data;
A step of selecting a process of acquiring a DC component from the orthogonal transform coefficient and generating 1 × 1 pixel data;
Combining the M row × M column pixel data obtained by the inverse orthogonal transform and the DC component, and smoothing the M row × M column pixel data as necessary to generate image data; When,
And a step of generating reduced image data of an arbitrary size from the generated image data. The image processing method according to claim 6, wherein, in the step of performing the inverse orthogonal transform, a value of the M rows × M columns is selected according to a reduction ratio of the image data. The image processing method according to claim 6, wherein, in the step of performing the inverse orthogonal transform, inverse transform of a plurality of sizes of the M rows × M columns is performed. The image processing method according to claim 6, wherein an edge portion in the block of the image data is extracted, and the smoothing is performed only on a block that is determined not to have the edge portion. The image processing method according to claim 6, wherein whether or not to perform the smoothing is selected based on a reduction rate of the image data and a quantization table value. In an image processing program for generating reduced image data from a compressed image obtained by orthogonally transforming image data from a spatial domain to a frequency domain,
M-row × M-column data (M is a natural number) of orthogonal transform coefficients and inverse orthogonal transform to generate M-row × M-column pixel data;
Selecting a process of obtaining a DC component of the orthogonal transform coefficient and generating 1 × 1 pixel data;
Combining the M row × M column pixel data obtained by the inverse orthogonal transform and the DC component, and smoothing the M row × M column pixel data as necessary to generate image data; When,
And a step of generating reduced image data of an arbitrary size from the generated image data. The image processing program according to claim 11, wherein in the step of performing the inverse orthogonal transform, a value of the M rows × M columns is selected according to a reduction rate of the image data. The image processing program according to claim 11, wherein the step of performing the inverse orthogonal transform performs an inverse transform of a plurality of sizes of the M rows × M columns. The image processing program according to claim 11, wherein an edge portion in the block of the image data is extracted, and the smoothing is performed only on a block that is determined not to have the edge portion. The image processing program according to claim 11, wherein whether to perform the smoothing is selected based on a reduction ratio of the image data and a quantization table value.

Images