JPH05284360A - Picture compression method and picture compressor - Google Patents

Abstract

PURPOSE:To set data quantity after compression within a specified value even when a natural picture and a color character or the like are in existence in mixture, to prevent deterioration in picture quality due to inadequate response to a compression table and to make data quantity after compression constant with less deterioration in picture quality. CONSTITUTION:Color picture data are converted into a spatial frequency component by an image area separation section 6, a characteristic of a color picture is identified based on the converted spatial frequency component, a quantization table 10 corresponding to the identified characteristic is selected, data quantity at compression is predicted by using the selected table and the picture is compressed by using a value of the quantization filter adjusted so that the predicted value is a specified value or below.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention uses an image compression method adapted to the characteristics of an input image to compress a multi-valued image within a prescribed data amount, and an image processing method and image for performing compression for improving the image quality. The present invention relates to a compression device.

[0002]

2. Description of the Related Art As a multivalued image compression technique, an ADCT (Adaptive Discrete Cosine Transform) compression method has been proposed so far mainly for natural images. In this compression method, a three-primary-color (RGB) signal is converted into three components of Y, U, and V, and a Y signal that is luminance has the same resolution and U, V of chromaticity components
The signal is compressed by reducing the resolution by sub-sampling as needed.

This conventional compression method will be described below. First, in step S1, for each component of Y, U and V, DCT is performed for each block of 8 × 8 pixels and converted into an 8 × 8 frequency space to obtain DCT coefficients. In the subsequent step S2, two kinds of quantization tables of the luminance component (Y) and the chromaticity component (U, V) are prepared, and the DCT coefficient obtained in step S1 is 8 × 8 quantization table for each component. 8 × 8, which is the element multiplied by the quantization factor
Is linearly quantized (divided) by the quantized value of, and the quantized coefficient is calculated. Then, in step S3, the quantized coefficient obtained in step S2 is Huffman-encoded.

[0004]

However, in the conventional compression method, when the compression processing is performed on the image in which the natural image, the color character image and the CG image are mixed, the Huffman coding unit of the ADCT compression method is variable. Since it is a long compression method, the amount of compressed data cannot be predicted. Therefore, it is necessary to prepare a large storage memory.

However, if the memory for the multi-valued color image is taken large, there is a drawback that the cost is high. Further, if the amount of compressed data is reduced in order to keep the memory amount within the specified value, there is a drawback that the image quality deteriorates significantly.

[0006]

The present invention has been made for the purpose of solving the above-mentioned problems, and has the following constitution as one means for solving the above-mentioned problems. That is, conversion means for converting color image data into spatial frequency components,
Feature identifying means for identifying features of the color image based on the spatial frequency components transformed by the transforming means, and respective quantization tables based on the results identified by the feature identifying means. Quantization table identifying means, and quantization factor selecting means for selecting a quantization factor in advance so that the amount of data compressed using the quantization table identified by the quantization table identifying means falls within a specified value.

[0007]

In the above construction, first, for example, prescan is performed before compression, and a quantization table adapted to each of several types of quantization tables is created by using a difference in frequency characteristic for each predetermined block, for example, 8 × 8 blocks. It is possible to predict the expected value of the total amount of data after compression when using each quantization table and determine the value of the quantization factor so that this expected value will be the specified amount of data. it can.

Therefore, if the value of the quantization factor is determined by the above configuration and then this scanning is performed, the data amount after compression is kept within the specified value even when a natural image, a color character, etc. are mixed. In addition, it is possible to prevent image quality deterioration due to incompatibility of the compression table, and it is possible to make the amount of data after compression constant in a state where the image quality deterioration is small.

[0009]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment according to the present invention will be described in detail below with reference to the drawings. [First Embodiment] In the embodiment described below, when compressing an image in which various frequency characteristics such as a natural image and a color character image are mixed, the compressed data is stored in a certain size, and the image quality is not deteriorated. Characterized by making few images.

FIG. 1 is a block diagram showing the configuration of an embodiment according to the present invention. In FIG. 1, 1 is a color conversion unit,
2 is a sub-sampling unit, 3 is a DCT unit, 4 is a quantization unit, 5 is a Huffman coding unit, 6 is an image area separation unit, 7 is a table switching switch, 8 is a quantization factor, and 9 is a quantization factor calculation unit. Reference numeral 10 is a quantization table, and in this embodiment, a total of 16 tables 1 to 16 having different characteristics are provided. Further, 11 is a code part, and 16 is a switch.

Image data compression processing of this embodiment having the above configuration will be described below. First, a prescan operation is performed prior to the actual document reading operation in order to set the quantum factor. Then, the factors etc. required for each quantization are determined, and then the main scan for encoding the original is performed. The pre-scan and main scan are, for example,
The image scanner is composed of a CCD sensor. At the time of prescan, the switch 16 is connected to the quantization factor calculator 9 side, and the signal from the image area separation unit 6 is sent to the quantization factor calculator 9. Further, the quantizing unit 4 is not activated, and the signal from the DCT unit 3 is sent to the image area separating unit 6 and is processed there.

RGB image data read by an image scanner (not shown) is input to the color conversion section 1. The color conversion unit 1 converts the RGB image data into a corresponding YUV signal according to the 3 × 3 linear matrix conversion shown in Expression 1 below. In the YUV signal converted by the color conversion unit 1, Y represents a luminance component and U and V represent chromaticity components.

In the sub-sampling unit 2, the sensitivity characteristic of the human eye is a luminance component (Y) rather than a chromaticity component (U, V).
Taking advantage of the fact that is more sensitive, subsampling is performed. Specifically, it is converted into Y: U: V = 4: 1: 1 or Y: U: V = 4: 2: 2. Then D
In the CT unit 3, 8 for each signal of Y, U and V
DCT (discrete cosine transform) for each × 8 block, 8 × 8
Frequency conversion is performed for each block. This frequency-converted coefficient is called a DCT coefficient.

The image area separation unit 6 examines the frequency characteristics of Y, U, V, and three types of DCT coefficients for each 8 × 8 block, identifies the image attribute in each block, and determines each table of the quantization factor 8. From 1 to 16, a table is selected so as to save the frequency characteristic for each specified 8 × 8 block. For example, when it is determined that the image attribute is a character portion, the character portion contains a large amount of high frequency components, and therefore a selection is made such as selecting a table in which the information of the high frequency components is not lost.

The quantizer factor calculation unit 9 sets 1 in the quantizer table 10 when the quantizer factor is set to a reference value.
The expected value (An) of the amount of data obtained by compressing 8 × 8 blocks is registered for each of the tables (16). Therefore, the expected value S of the data amount when the whole is compressed is obtained from the number (In) of the table numbers in the quantization table 10 selected by the image area separation unit 6 using the following Expression 2.

S = Σ (In × An) (Equation 2) where In is the number using the table n and An is 8 × 8
Expected value of the amount of data when the block is compressed by the table n, S is the expected value of the amount of data of the entire compressed image expected when the quantization factor is 1, The expected value S of the data amount at the time of performing is the value when the quantization factor is the reference value. Therefore, the quantization factor is empirically reset so that S is within the specified value of data.

This quantization factor keeps a constant value during compression and does not change the compression rate by changing the size. Therefore, in this embodiment, a factor having a margin is determined as shown in the following expression 3 so that the expected value S is always within the specified value. Data expected value ≧ S (Equation 3) In this way, in the prescan, the table used for quantization in the quantization table 10 is specified, and the quantization factor is set.

The functions of the image area separation unit 9, the quantization factor calculation unit 9 and the like in the above-described prescan operation are slightly different from those in the present scan, and when the functions are shown as a block structure, the block structure shown in FIG. 2 is obtained. That is, as shown in FIG. 2, the function of examining the frequency characteristics of the three types of DCT coefficients of Y, U, V by the image area separating unit 6 is the frequency characteristic analyzing unit 13
The function of specifying the image attribute in each block, that is, the function of specifying the parameter for performing the image area separation can be expressed by the image area separation parameter calculation unit 14.

Further, the function for specifying the table used for quantization in the quantization table 10 by the quantization factor calculator 9 can be represented by the quantization table calculator 15. It should be noted that the above-described functions are provided as independent functions separated from the functions achieved during the main scanning, and the configuration shown in FIG. 2 is used during the pre-scanning, and the output from the DCT unit 3 is configured using the configuration shown in FIG. 1 during the main scanning. You may make it output to 4 etc.

In the subsequent scanning, the quantizer 4 is activated and the switch 16 is switched to the table 7
The control signal is switched to the side and the select signal from the image area separation unit 6 is controlled to be sent to the table switching switch 7. And
Perform the same document reading process as in the case of prescan.
The RGB image data read from the original is converted into a corresponding YUV signal by the color conversion unit 1, subjected to sub-BPU ring by the sub-sampling unit 2, and sent to the DCT unit 3. Then, DCT is performed by the DCT unit 3 and sent to the image area separation unit 6 and the quantization unit 4.

The quantizing unit 4 uses the selected quantizing table and the quantizing factor obtained by the prescanning by the table switching switch 7 for each of Y, U, and V signals at every 8 × 8 blocks. Quantize. The quantized coefficient quantized by the quantizer 4 is Huffman-encoded by the Huffman encoder 5 and sent to the decompression side as compressed data.

It should be noted that the image area separation unit 6 selects a signal (4
The bit) is also sent to the encoding unit 11. Then, this select signal is encoded by the encoding unit 11 and is delivered to the decompression side as compressed table data. Further, the table row data of the quantization factor 8 and the quantization table 10 are also sent to the encoding unit 11 via the table switching switch 7, and these data are also encoded (compressed) and passed to the decompression side.

The coding by the coding unit 11 can be performed by any coding method such as MH coding, MR coding, MMR coding and the like. As a result, a quantization table suitable for blocks having different frequency characteristics can be set. Regarding the expansion of the compressed data compressed as described above, the image area separation unit 6 and the quantization factor calculation unit 9
You can remove it and reverse the signal flow. However, the Huffman coding unit 5 is replaced by a Huffman decoding unit, the quantizing unit 4 becomes an inverse quantizing unit, the DCT unit 3 becomes an inverse DCT unit, and the encoding unit 11 becomes a decoding unit.

In the sub-sampling unit 2, when the sub-sampling ratio is 4: 1: 1, Y1, Y2, Y3.
Y4, U1, V1 ... to Y1, Y2, Y3, Y4, U1,
Convert to U1, U1, U1, V1, V1, V1, V1 ... On the other hand, in the sub-sampling unit 2, when the sub-sampling ratio is 4: 2: 2, Y1, Y2, U
, 1, V1, ... Are converted to Y1, Y2, U1, U1, V1, V1.

The color conversion section 1 may perform the inverse conversion of (Equation 1). As the quantization table and the quantization factor, those passed from the compression side are used. The transmitted table data is decoded by the decoding unit, and the table switching switch 7 selects the table corresponding to the table at the time of compression by using this as a select signal. Dequantization is then performed based on the data from the selected table.

As described above, according to the present embodiment, first, the press scan is performed before the compression, and the difference in the frequency characteristic for each predetermined block, for example, 8 × 8 blocks is used.
Select a quantization table suitable for each of several types of quantization tables, predict the expected value of the total amount of data after compression when using each quantization table, and set this expected value to the specified amount of data. The value of the quantization factor can be determined so that If this scanning is performed after determining the value of the quantization factor in this way, the amount of data after compression is kept within the specified value even if natural images, color characters, etc. are mixed, and image quality deterioration due to compression table inadequacy Can be prevented, and the amount of data after compression can be made constant with little deterioration in image quality.

[Other Embodiments] The above description has explained the configuration in which the select signal obtained by the prescan is compressed by the encoding unit 11 and sent to the decompression side. However, the present invention is not limited to the above example, and may have a configuration including a table buffer for storing the select signal obtained by the prescan. FIG. 3 shows a second embodiment according to the present invention, which is provided with a table buffer for storing the select signal obtained by the prescan. In FIG. 3, the same components as those in the first embodiment shown in FIG. 1 described above are designated by the same reference numerals and detailed description thereof will be omitted.

In the second embodiment, a table buffer 12 is provided between the image area separating section 16 and the table switching switch 7 of FIG.
The configuration is such that the synchronization signal from the CT unit 3 is input. Then, the image area separation unit 6 and the quantization factor calculation unit 9 are controlled to operate only during prescan. That is, in this scanning, the image area separation unit 6 is not used, and the select signal stocked in the table buffer 12 is set to DC.
It is sent to the table switching switch 7 in synchronization with the synchronization signal from the T section 3.

According to the second embodiment having the above construction,
When the image area separation unit 6 requires a long processing time,
The execution time at the time of this scan can be shortened. As mentioned above,
By separating the image area and adaptively selecting the quantization table, it is possible to prevent the deterioration of the image in which characters and natural images are mixed, and by adjusting the quantization factor, the compressed data The size can be adjusted.

In the embodiment described above, the quantization table 10 has 16 types of 1 to 16, but
Other numbers such as 17 types may be used. Although the select signal is encoded by the encoding unit 11, it may be directly passed to the decompression unit as table data without being encoded. Further, in the above embodiment, in order to obtain the expected value S of the data amount when the whole is compressed, the expected value An of the compressed data amount of 8 × 8 blocks for each table when the quantization factor is used as the reference value An. The example of obtaining by (Equation 3) has been described using. However, the present invention is not limited to the above example, and the expected value An (q) of the compressed data for each 8 × 8 block is previously set for the combination of N quantization tables and Q quantization factors. It may be configured to be registered. Then, the quantization table is determined for each block based on the image area separation, and the compressed data expected value S (q) of the entire image when the quantization factor is changed is calculated from the following equation (4), A configuration may be used in which the scanning is performed by using a quantization factor in which S (q) is equal to or less than a target specified value and is maximum.

S (q) = Σ (In × An (q)) (Equation 4) where An (q) is an 8 × 8 block using (quantization table number n) and (quantization factor number q). Is the expected value of the compressed data amount. As described above, by separating the image area and adaptively selecting the quantization table, it is possible to prevent deterioration of the image in which characters and natural images are mixed, and by adjusting the quantization factor, The size of the compressed data can be adjusted.

[0032]

As described above, according to the present invention, first, for example, pre-scan is performed before compression, and by applying the difference in the frequency characteristic for each predetermined block, it is adapted to each of several kinds of quantization tables. Select a quantization table, predict the expected value of the total data amount after compression when using each quantization table, and set the value of the quantization factor so that this expected value will be the specified data amount. You can decide.

Therefore, if the value of the quantization factor is determined by the above-mentioned configuration and then this scanning is performed, the data amount after compression is kept within the specified value even when a natural image, a color character, etc. are mixed. In addition, it is possible to prevent image quality deterioration due to incompatibility of the compression table, and it is possible to make the amount of data after compression constant in a state where the image quality deterioration is small.

[Brief description of drawings]

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

FIG. 2 is a block diagram for calculating an image separation parameter based on prescan data and a quantization table in the present embodiment.

FIG. 3 is a block diagram showing the configuration of another embodiment according to the present invention.

[Explanation of symbols]

 1 Color conversion unit 2 Sub-sampling unit 3 DCT unit 4 Quantization unit 5 Huffman coding unit 6 Image area separation unit 7 Table switching switch 8 Quantization factor 9 Quantization factor calculation unit 10 Quantization table (no. 1 to no. 16) 11 coding unit 12 table buffer 13 frequency characteristic analysis unit 14 image area separation parameter 15 quantization table calculation unit 16 switch

Claims (2)

[Claims] 1. Color image data is converted into spatial frequency components, features of the color image are identified based on the transformed spatial frequency components, and compression is performed using a quantization table corresponding to the identified features. An image compression method characterized in that the amount of data is predicted in advance, and the image is compressed using a value obtained by adjusting the quantization filter so that the predicted value becomes equal to or less than a specified value. 2. A conversion unit for converting color image data into a spatial frequency component, a characteristic identification unit for identifying a characteristic of the color image based on the spatial frequency component converted by the conversion unit, and the characteristic identification. A quantization table identifying means for identifying each quantization table based on the result identified by the means, and a data amount compressed using the quantization table identified by the quantization table identifying means within a specified value. An image compression apparatus comprising: a quantization factor selecting means for selecting a quantization factor in advance so as to be stored.