JP2003116130A - Image processing apparatus and image compression method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an image processing apparatus capable of generating compression data comprising a code quantity with high accuracy with respect to a target code quantity by deciding an SF (Scale Factor) value by taking into account a code quantity characteristic of image data. SOLUTION: A discrimination section 12 discriminates the code quantity characteristic of image data on the basis of a generated code quantity obtained by initial coding processing. A selection section 11 selects an SF table corresponding to the discriminated code quantity characteristic from a table group 13 and sets an SF value adapted to the target code quantity from the SF table to a quantization processing section of a compression processing section 2.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention generally relates to a digital image compression system, and more particularly to an image compression system in which the code amount characteristic of image data is taken into consideration.

[0002]

2. Description of the Related Art Conventionally, for example, in a digital camera,
It is common to compress a still image (a captured image with a camera) and record it on a recording medium. As an image compression method,
A JPEG compression method based on the so-called JPEG standard is well known. The process of image compression processing by the JPEG compression method will be described below with reference to FIG.

For example, in a digital camera, the image pickup system 30 generates digital image data (hereinafter simply referred to as image data) corresponding to a photographed subject. The compression processing unit 2 uses, for example, an ASIC (Application Specific Integ
The image data (including the YC signal) input from the imaging system 30 is subjected to a series of compression processes. First, the image data is usually divided into blocks each including 8 × 8 pixels as a block unit. The DCT (discrete cosine transform) calculation unit 21 executes a DCT calculation process for each block to calculate a DCT coefficient. The quantization processing unit 22 executes a quantization (linear quantization) process on the DCT coefficient according to each frequency component.
Furthermore, the encoding processing unit 23 normally performs entropy encoding processing (hereinafter simply referred to as encoding processing) using Huffman code on the quantized value. The output unit 24 accumulates the compressed data of the code amount by the encoding process and outputs it to the recording system 3 using a recording medium such as a memory card.

In such a compression process, the quantization processing unit 22 normally uses a quantization table for determining the compression rate and executes the quantization process so as to relatively reduce the code amount of the high frequency component. To do. Images of high-frequency components are difficult to detect visually, so even if they are reduced, there is little deterioration in image quality.
Practically, an optimum quantization table obtained by multiplying a default quantization table 42 prepared in advance with a coefficient (hereinafter sometimes referred to as SF) 41 called a scale factor is used. Specifically, the quantization table calculation unit 40 executes a multiplication process to calculate an optimum quantization table value.

In short, by adjusting the SF, the quantization process can be adjusted to change the compression rate of the image data. Generally, for the same image data,
The larger the SF, the larger the compression rate.

[0006]

As described above, in the JPEG compression method, for example, in the process in the quantization processing unit 22, the compression rate (in other words, the target code amount) of the image data is set in accordance with the SF adjustment. Can be adjusted.

As a prior art, there has been proposed a method for obtaining an optimum code amount by utilizing the relationship between the code amount and SF (see, for example, Japanese Patent Laid-Open No. 2000-114980). Specifically, the first compression process is executed using the default SF, and the relational expression (1) between the generated code amount and the SF
From S to be used in the next compression process that matches the target code amount
This is a method of determining F.

LogBR = a × LogSF + b (1) Here, BR means the code amount, and a and b are parameters that differ depending on the image. However, the parameter a
Is a substantially constant value regardless of the image if the same encoding is performed. The parameter b is a value that depends on the image.

In short, the method of the prior art performs compression processing on various images in advance and obtains the parameters a and b by statistical processing. Then, the SF value used for the next compression process is determined from the SF value used for the first compression process and the comparison result between the target code amount and the generated code amount (the code amount obtained by the compression process). ing.

However, the prior art methods do not take into account the characteristics or types of captured images, for example digital cameras. The features or types of images, which should also be referred to as code amount characteristics, are generally relatively complicated images (for example, images of trees with many branches) and monotonous images (for example, distant view images). Is roughly divided into If such a code amount characteristic is not taken into consideration, the value of the SF used in the compression process cannot be accurately determined, so the variation in the code amount generated according to the image becomes extremely large.

Therefore, an object of the present invention is to enable determination of the SF value in consideration of the code amount characteristic of image data,
An object of the present invention is to provide an image processing device capable of generating compressed data having a highly accurate code amount with respect to a target code amount.

[0012]

SUMMARY OF THE INVENTION An aspect of the present invention is that in a still image compression method (for example, JPEG standard method), D
The present invention relates to an image compression method that realizes means for setting an SF value (SF that matches a target code amount) in consideration of a code amount characteristic of image data in a quantization process after CT processing.

The code amount characteristic of image data corresponds to the feature or type of the image, and specifically, it is a complex image (an image generally having many high frequency components) and a monotonous image (generally a low frequency component). There are many images). As a method of determining the code amount characteristic, the determination is performed based on the generated code amount after the encoding process for the same SF.

An image processing apparatus adopting the system of the present invention is
By executing the quantization process and the encoding process,
In an image processing apparatus for compressing image data, a storage unit that stores table information in which a scale factor, which is a coefficient of a quantization table used in the quantization process, and a code amount characteristic of the image data are stored, and an initial encoding process. Determining means for determining the code amount characteristic of the image data based on the code amount obtained by, and selecting means for selecting a scale factor corresponding to the code amount characteristic determined by the determining means from the table information, The scale factor selected by the means is used to perform the encoding process based on the quantized value obtained by performing the quantization process again.

With such a configuration, in the process of the quantization process, when setting the value of the SF that matches the target code amount, the setting can be performed in consideration of the code amount characteristic of the image data. Therefore, as a result, it is possible to generate compressed data having a highly accurate code amount with respect to the target code amount.

[0016]

BEST MODE FOR CARRYING OUT THE INVENTION Embodiments of the present invention will be described below with reference to the drawings.

(System Configuration) The image processing apparatus of the embodiment executes a still image compression process corresponding to a so-called JPEG compression method, and is adopted as an image processing device such as a digital camera.

As shown in FIG. 1, the image processing apparatus of the embodiment has a buffer memory 1 storing image data before compression, a compression processing section 2, a recording system 3, and an SF value setting section 4. Have and. The buffer memory 1 is a memory for temporarily storing image data output from the image pickup system (30) of the digital camera, for example. The compression processing unit 2 is composed of, for example, components according to the JPEG compression method, and the DCT
It has an operation unit, a quantization processing unit, an encoding processing unit, an output unit, a default quantization table, and a quantization table operation unit (see FIG. 5).

The SF value setting unit 4 of the embodiment has a scale factor value to be multiplied by the default quantization table in order to calculate an optimum quantization table value by the quantization table operation unit included in the compression processing unit 2. Is a processing unit for setting.

The SF value setting unit 4 uses an initial SF value (here, a fixed value of 0.
02) is stored, a selection unit 11, a determination unit 12, an SF table group 13, and a memory 14 holding evaluation information. The discrimination unit 12 discriminates the code amount characteristic (corresponding to the feature or type of the image) of the image data from the generated code amount obtained by the initial encoding process based on the evaluation information (see FIG. 3). The selection unit 11 includes the determination unit 12
According to the determination result from, the SF table that matches the code amount characteristic is selected from the SF table group 13, the SF value that matches the target code amount is determined from the SF table, and the SF value is output to the compression processing unit 2.

(Image compression process) Referring to FIG. 1 and the flowchart of FIG. 2, FIG. 3 and FIG.
The image compression process of the embodiment will be described.

The process of the first embodiment includes an initial coding process using the initial SF value (first coding process) and an encoding process using the set SF value (second coding process). It is roughly divided into. First, the compression processing unit 2 normally takes 8 × for image data (including YC signal) that is a captured image obtained by a capturing operation of a digital camera, for example.
Block processing is performed to divide the block into blocks each having 8 pixels (step S1). Further, the compression processing unit 2 executes the DCT calculation process for each block to perform the DCT.
The coefficient is calculated (step S2).

Next, the quantization processing unit included in the compression processing unit 2 executes a quantization (linear quantization) process on the DCT coefficient according to each frequency component. Here, the quantization processing unit includes a quantization table calculation unit 40 and a default quantization table 42. The quantization processing unit executes the quantization processing using the initial SF value (fixed value 0.02) stored in advance in the memory 10 (step S3).

The encoding processing unit included in the compression processing unit 2 is
An initial coding process (entropy coding process by Huffman coding) is performed on the quantized value using the initial SF value (step S4). The output unit 24 outputs the generated code amount by the initial encoding process to the determination unit 12 of the SF value setting unit 4.

The discriminating section 12 discriminates the code amount characteristic of the image data by using the evaluation information stored in the memory 14. As shown in FIG. 3, the evaluation information is obtained by previously evaluating the generated code amount for each of the image types 30, 31, and 32 with the SF value as a reference. Here, the same SF value (initial S
A complex image (image having many high frequency components) 30, a relatively complex image 31, and a relatively monotonous image (low frequency) with a relatively large generated code amount with respect to the F value of 0.02). Image with many components) 32.

It is assumed that the discriminating unit 12 discriminates that the image data is the complex image 30 based on the generated code amount and the evaluation information from the compression processing unit 2, for example. The selection unit 11 selects a compatible SF table from the SF table group 13 according to the determination result of the determination unit 12 (step S6).

As shown in FIG. 4, the SF table group 13 has a code amount characteristic of image data, that is, image types 30, 3
It has respective tables corresponding to 1 and 32, respectively. Here, the table corresponding to the complicated image (image having many high frequency components) 30 is table A as shown in FIG. The table corresponding to the moderately complex image 31 is table B, as shown in FIG. Further, the table corresponding to the monotonous image (image having many low frequency components) 32 is table C as shown in FIG.

The selecting unit 11 selects the table A corresponding to the complicated image 30 from the SF table group 13 according to the discrimination result of the discriminating unit 12, and according to the target code amount (that is, the compression rate) from the table A. The SF value is extracted and set in the quantization processing unit of the compression processing unit 2. As a result, the quantization processing unit executes the quantization processing using the set SF value (a value considering the code amount characteristic) (step S).
7). The encoding processing unit included in the compression processing unit 2 executes the second encoding process on the quantized value from the quantization processing unit (step S8). Here, the compression processing unit 2
The DCT coefficient calculated during the initial encoding process is stored, and the second quantization process is executed on the DCT coefficient (step S7). However, the compression processing unit 2 may re-execute the blocking process and the DCT process before starting the second quantization process.

The output unit 24 accumulates the compressed data of the code amount by the encoding process and outputs it to the recording system 3 using a recording medium such as a memory card (step S9).

When the discrimination result of the discriminating unit 12 is an image 31 having a medium complexity, the selecting unit 11 selects the SF table group 13
The table B is selected from the table B, the SF value corresponding to the target code amount is extracted from the table B, and is set in the quantization processing unit of the compression processing unit 2. Further, when the determination result of the determination unit 12 is the monotonous image 32, the selection unit 11 selects the table C from the SF table group 13 and extracts the SF value according to the target code amount from the table C, It is set in the quantization processing unit of the compression processing unit 2.

In summary, according to the method of the embodiment, in the process of setting the SF value by the initial coding process, the discriminating unit 12 causes the code amount characteristic of the image data (image type 30
32 to 32) are discriminated, and a suitable SF table is selected from the discrimination result. Therefore, as a result, the SF value considering the code amount characteristic of the image data can be set in the quantization processing unit of the compression processing unit 2. As a result, it is possible to suppress variations in the generated code amount depending on the type of image and generate a highly accurate code amount close to the target code amount.

By the way, even if the compression processing unit 2 executes the second encoding process, there is a possibility that the generated code amount does not fall within the allowable range of the target code amount (NO in step S10). . In this case, generally, the SF table cannot be used again to obtain an appropriate SF value.
Therefore, the above-described relational expression (1) between the generated code amount and SF is used to substitute the values of the SF value and the generated code amount for the first and second times to calculate the parameters a and b. Then, the parameters a and b are substituted into the relational expression (1) to calculate the SF value that matches the target code amount (step S11). The quantization processing unit of the compression processing unit 2 executes the quantization processing using the SF value. Furthermore, the encoding processing unit included in the compression processing unit 2 sets the quantization value from the quantization processing unit to 3
The encoding process for the second time is executed (step S12).

(Modification) In the embodiment, as the code amount characteristics of the image data, a complex image (image having many high frequency components) 30, a relatively large generated code amount, a medium complex image 31, and a relative image Although there are three types of images that are monotonous (images with many low frequency components) 32, the present invention is not limited to this. As a modified example, for example, five or more types may be used. However, SF
The table group 13 has five images corresponding to the assumed five types of images.
You will need a table. This makes it possible to obtain a higher-accuracy generated code amount in proportion to an increase in the classification of code amount characteristics of image data.

(Modification) In the same embodiment, the resolution of each table of the SF table group 13 is the same. However, it is generally difficult to approximate the above relational expression (1) in the case of a code amount characteristic in which the generated code amount is small (an image having many low frequency components). Therefore, it is desirable that the tables B and C corresponding to the code amount characteristic in which the generated code amount is relatively small have a data group having a relatively high resolution. This makes it possible to obtain a highly accurate generated code amount even when the generated code amount has a relatively small code amount characteristic.

[0035]

As described above in detail, according to the present invention, the SF value can be set in the image compression processing process in consideration of the code amount characteristic of the image data corresponding to the feature or type of the image captured by the digital camera. . Therefore, it is possible to suppress the variation in the generated code amount due to the code amount characteristic and generate the image compressed data having a highly accurate code amount with respect to the target code amount. In other words, complex images with many high frequency components,
An appropriate SF value can be set according to the type of image such as a monotone image having many low frequency components. This makes it possible to realize highly accurate and efficient image compression processing.

[Brief description of drawings]

FIG. 1 is a block diagram showing a configuration of an image processing apparatus according to an embodiment of the present invention.

FIG. 2 is a flowchart for explaining a process of image compression processing according to the same embodiment.

FIG. 3 is a diagram for explaining an operation of a determination unit according to the same embodiment.

FIG. 4 is an exemplary view for explaining a configuration of a quantization table according to the same embodiment.

FIG. 5 is a block diagram for explaining a processing process according to a conventional image compression method.

[Explanation of symbols]

1 ... Buffer memory 2 ... Compression processing unit 3 ... Recording system 4 ... SF value setting section 10 ... Memory (initial SF value) 11 ... Selector 12 ... Discrimination unit 13 ... SF table group 14 ... Memory (evaluation information) 21 ... DCT calculator 22 ... Quantization processing unit 23 ... Encoding processing unit 24 ... Output unit 40 ... Quantization table operation unit 42 ... Default quantization table

Continued front page    F-term (reference) 5C022 AA13                 5C059 MA00 MA23 MC14 MC38 ME02                       PP01 SS15 TA47 TB04 TC10                       TC18 TC24 TC38 TD01 TD11                       TD15 TD16 TD17 UA02 UA38                 5C078 AA04 BA57 CA01 DA01 DA07                       DA11 DB07                 5J064 AA01 BA16 BC01 BC14 BC16                       BC25 BC29 BD03

Claims (7)

[Claims] 1. In an image processing apparatus for compressing image data by executing each process of a quantization process and an encoding process, a scale factor which is a coefficient of a quantization table used in the quantization process and a code of the image data. Storage means for storing table information in which the quantity characteristics are associated with each other; discrimination means for discriminating the code quantity characteristic of the image data based on the code quantity obtained by the initial encoding processing; and the discrimination means. A scale factor corresponding to the code amount characteristic, selecting means for selecting from the table information, and using the scale factor selected by the selecting means, based on the quantized value obtained by performing the quantization process again. An image processing apparatus characterized by being configured to perform an encoding process. 2. The initial coding process is adapted to code a quantized value obtained by a quantizing process using a scale factor prepared in advance as an initial value,
The image processing apparatus according to claim 1, wherein after the encoding processing, the encoding processing is performed again based on a quantized value obtained by the quantization processing using the selected scale factor. 3. The table information includes a plurality of tables for each code amount characteristic, and each table has a data group in which a target code amount for compression processing and a scale factor are associated with each other. The image processing device according to claim 1. 4. The discrimination means prepares an evaluation value indicating a code amount corresponding to a predetermined same scale factor for each code amount characteristic of image data, and the initial encoding using the same scale factor as an initial value. The image processing apparatus according to claim 1, wherein the image processing apparatus is configured to determine a code amount characteristic of image data based on a result of evaluation of a code amount obtained by processing with the evaluation value. 5. The image processing apparatus according to claim 3, wherein each of the tables has a data group set with a different resolution for each code amount characteristic. 6. The table of the code amount characteristics in which the code amount corresponding to the same scale factor is relatively small among the tables has a data group showing relatively high resolution. The image processing device according to item 3. 7. An image compression method for compressing image data by executing each process of a quantization process and an encoding process, the scale factor being a coefficient of a quantization table used in the quantization process and the image data. Preparing table information in which the code amount characteristics are associated with each other, a step of executing an initial encoding process of encoding a quantized value obtained by a quantization process using a scale factor prepared in advance as an initial value, A step of determining the code amount characteristic of the image data based on the code amount obtained by the initial encoding processing step, and a step of selecting a scale factor corresponding to the code amount characteristic determined by the determining step from the table information And a step of performing a quantization process using the scale factor selected by the selecting step, Image compression method characterized by comprising a step of processing the encoded quantized values obtained by the quantization processing steps.