JP2002374533A - Image coder, image decoder and method therefor - Google Patents

Abstract

PROBLEM TO BE SOLVED: To warrant both an edge position specified with a perceptible resolution and a perceptible number of gray scales and to code an image at high efficiency. SOLUTION: An input image 101 is given to a quantization LUT 102 and a reduction circuit 105. The quantization LUT 102 converts e.g. 8-bit data into 4-bit data. The reduction circuit 105 sums all pixel values in e.g. 3×3 blocks and divides the sum by 9 to convert the input image 101 into a reduced image whose longitudinal and lateral sizes are reduced 1/3 respectively. A 1st coder 103 encodes a quantization index quantized by the quantization LUT 102 into a 1st code 104. A 2nd coder 105 encodes the reduced image reduced by the reduction circuit 105 into a 2nd code 107.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image coding apparatus for compressing image data to reduce the amount of data.

[0002]

2. Description of the Related Art In order to obtain a high-quality digital image, a larger number of gradations and a higher resolution are required. The capacity of an image is represented by (the number of pixels × the number of gradation bits) and is enormous. Therefore, image compression is performed for the purpose of reducing the image storage cost or the image transmission time.

Conventionally, an image compression method using human visual characteristics has been used. As an example of visual characteristics, reference 1
(Peter GJ Barten, "Evalua
Tion of subjective image
quality with the square-ro
ot integral method, "J. Op.
t. Soc. Am. A / Vol. 7, No. 10 / Oc
t. 1990, p. 2024-2031).
This document shows an approximate expression of visual frequency characteristics.
FIG. 13 shows a graph of the approximate expression of the document. The horizontal axis of this graph is the number of cycles per 1 degree of viewing angle, and the vertical axis is the contrast sensitivity. According to this graph, the contrast sensitivity decreases in the high frequency portion. In other words, it indicates that a high frequency portion cannot be perceived even if the number of gradations is reduced. A compression method utilizing this property has been conventionally performed.

An example of a compression method using visual characteristics is described below. There are two examples below. The first example is a method of performing compression by directly reducing the number of pixels or the number of gradations of an image (conventional example 1-1 and conventional example 1).
-2). The second example is a method in which an input image is orthogonally transformed, and a quantization method of a transform coefficient is matched with a visual characteristic (conventional example 2).

First, a first example (conventional example 1-1, conventional example 1)
-2) will be described.

The capacity of an image is (number of pixels × number of gradation bits)
Therefore, if the number of pixels is reduced or the number of gradation bits is reduced, image compression can be performed. A portion such as a character or a line drawing in an image is a portion having many high frequency components, and therefore does not require the number of gradations. This part can be compressed by reducing the number of gradation bits.

When the frequency component of the input image is low, the number of gradations cannot be reduced. However, since the frequency band of the image is small, the image can be reduced without aliasing. In other words, a natural image portion such as a photograph in an image is a portion having a small number of high-frequency components. Therefore, this portion can be compressed by reducing the number of pixels.

[0008] In summary, low frequency parts can be compressed by reducing the number of pixels. High frequency portions can be compressed by reducing the number of gradations.

As an example of the above system, Japanese Patent Publication No. 6-4
No. 0661 can be mentioned. Hereinafter, the method of Japanese Patent Publication No. 6-40661 is referred to as Conventional Example 1-1. In the conventional example 1-1, an input image of 6 bits per pixel is divided into 3 × 3 pixel blocks. Next, it is determined whether this pixel block is coded in multi-value or binary. As shown in FIG. 11, when encoding with multi-values,
The average value of 9 pixels in the 3 × 3 block is calculated, and is set as 6-bit 1-pixel information. When coding in binary, 3 × 3
Each of the 9 pixels in the block is binarized, and 1-bit 9
This is pixel information. The information of the 3 × 3 block is 6 bits 1
The information is pixel information or 1-bit 9-pixel information.
At the time of encoding, one bit of information indicating whether the block is a multi-level block or a binary block is added to each block. In summary, 3 × 3
Encoding can be performed with a worst 10 bits per block of pixels.

As still another example, Japanese Patent Application Laid-Open No. H8-3172
The method disclosed in Japanese Patent Publication No. 20 will be described. Hereinafter, JP-A-8-317
No. 220 is referred to as Conventional Example 1-2. In the conventional example 1-2, for example, an input image is divided into 2 × 2 pixel blocks.
Next, it is determined whether this pixel block is coded in multi-value or binary. When encoding with multiple values, 2 ×
The average value of four pixels in two blocks is encoded. In the case of binary coding, the number of black pixels in a 2 × 2 pixel block is coded. As an example, the average value level number in the case of multi-level encoding is quantized into 12 levels of 0 to 11. Also,
The number of black pixels for binary coding is 0, 1, 2, 3,
4. When the number of black pixels is 0, it represents the same state as the multi-level level 0. When the number of black pixels is 4, it represents the same state as the multi-level level 11. Therefore, as shown in FIG. 12, it is possible to perform encoding at 15 levels in both the case of multi-value and the case of binary.

The conventional example 1-2 is different from the conventional example 1-1 in that the codes for the multi-valued case and the binary case are made to have the same code space so that the multi-valued / binary discrimination bit is unnecessary. .
However, the same technique is used for a configuration in which encoding is performed by discriminating whether to perform multi-level encoding or binary encoding for each block.

As described above, in the conventional example 1-1 and the conventional example 1-2, it is determined whether the input image block is low-frequency information requiring the number of gradations or high-frequency information not requiring the number of gradations. Determine. Next, in the case of low frequency information requiring the number of gradations, the average value of the pixel values in the block is encoded. In the case of high-frequency information that does not require the number of gradations, the compression ratio is increased by reducing the number of gradations of the pixel values in the block (in the above example, the number of gradations is binary).

Next, a second compression method using visual characteristics is described.
(Conventional example 2) is shown below.

At present, the still image coding system is a still image coding international standard JPEG system (ITU-T Recommendation T.8).
1) The method is most widely used. As a document explaining the outline of the JPEG system, “International Standard for Multimedia Coding (ed by Yasuda, Maruzen), pp. 16-47” can be mentioned. Hereinafter, the JPEG method is referred to as Conventional Example 2.

In the JPEG system, an input image is divided into blocks of 8 × 8 pixels, and DCT (discrete cosine transform) is performed for each block to obtain 8 × 8 DCT coefficients. Furthermore,
Compression is performed by quantizing 8 × 8 DCT coefficients. 8.times.8 quantization step widths (quantization tables) are not specified in the standard and can be set freely. Therefore, by designing a quantization table in consideration of visual characteristics, encoding with a higher compression rate can be realized.

For example, as a design example of a quantization table, a document (Saito et al., "Optimal quantization of DCT coefficients in consideration of visual characteristics in variable-length coding," IEICE Technical Report IE90
-101). This document is DC
In an encoding method using T (discrete cosine transform),
This describes a method of designing a quantization table in consideration of visual characteristics. Since the JPEG method is a method using DCT, the method described in this document is a JPE method.
It can also be used when designing a G-system quantization table. The quantization step width (quantization table) of each of the 8 × 8 DCT coefficients is designed so that visual distortion is minimized for a given code amount in consideration of visual characteristics. The visual characteristics considered here are visual frequency characteristics similar to those shown in Reference 1. Here, the visual frequency characteristics

H (f) = 2.46 (0.1 + 0.25f) e
xp (−0.25f). Here, f in the above equation is the number of cycles per unit viewing angle (cycle / degree).
It is.

[0017]

However, the above-mentioned prior art has the following problems.

(1) Problem 1: It is not possible to cope with a case where a portion requiring a resolution and a portion requiring a number of gradations are mixed in a block. (2) Problem 2: When the gradation level indicates the edge position, image distortion occurs.

Hereinafter, Problem 1 and Problem 2 will be described in detail.

The problem 1 will be described below.

As shown in FIG. 14, consider a block in which a character or a line drawing is overwritten with a photograph as a background. A photograph is data that requires a number of gradations, and a character or a line drawing is data that requires a resolution. In the case of such a block, if the block is determined as the multi-value information and the average value of the block is encoded, the resolution information of the character and line drawing is lost. Further, if a block is determined as a binary (or low-value) block and the pixel values in the block are quantized, the number of gradations of the photograph is lost. In order to avoid this, the block size of the discrimination unit must be reduced. The worst case is one pixel per block. By reducing the block size, the compression ratio decreases. That is, Conventional Example 1-1 and Conventional Example 1-2
Cannot cope with a case where a portion requiring a resolution and a portion requiring a gradation number are mixed in a block.

The problem 2 will be described below.

As an example of the case where the gradation level indicates the edge position, there is a gray font. Usually, an imager, which is software for generating an image for a printer, generates a character image as a binary image. In this case, characters cannot be represented with a resolution higher than the pixel resolution. FIG.
Shows an image generated in binary. The rectangle in FIG. 15 corresponds to one pixel. As shown in FIG. 15, jaggies occur in oblique lines. Alternatively, there is a problem that the line width and the edge position are limited to an integral multiple of the pixel resolution. Therefore, a gray font is designed to improve the character image quality without increasing the pixel resolution by generating a character image in multiple values. As shown in FIG. 16, pixels are represented by multi-values. When looking at the pixels represented by multi-values, they are complemented by the visual characteristics (or by controlling the marking position according to the pixel values), other than diagonal lines as shown in the lower part of FIG. The width and position of the line can be represented. As described above, in the gray font, the pixel value is information for controlling the position of the edge. The gray font is an example in which a multi-gradation image is artificially generated so that the gradation level indicates an edge position. Such an image exists even if it is an image of the natural world. It is considered that the same effect as that of the gray font artificially generated can be obtained even when the character image is scanned in with multi-values. Further, when there is an edge in the input image, it is considered that the gradation level has the effect of representing the edge position in the same manner, though there is a difference in the degree of influence on vision.

In the case of such an image in which the number of gradations indicates the edge position, if the number of gradations is changed, the edge position moves, and is perceived as image distortion. In the literature (Oyama et al., "New Edition Handbook of Sensory and Perceptual Psychology @ Seishin Shobo, pp. 557-558"), a perceptual threshold value relating to the edge position is indicated by the name of vernier visual acuity. As shown in FIG. 17, the minimum deviation that can distinguish the deviation between the two straight lines, that is, the threshold value of the vernier visual acuity is 2 ″ in the viewing angle. The deviation of the edge position must be suppressed within a viewing angle of 2 ″.

In the case of Conventional Examples 1-1 and 1-2, if a block is determined to be multi-valued, the resolution is changed to a lower resolution than the pixel resolution. At this time, edge position information higher than the pixel resolution described above is lost. If the block is determined to be binary (small),
Since the pixels are quantized, an error occurs at the edge position. In either case, image distortion occurs.

In the case of the conventional example 2, since the quantization is performed in the orthogonal transform domain, it is difficult to guarantee the error amount of each pixel value. Although there has been a quantization table design method in consideration of visual frequency characteristics, there is no quantization table design method that guarantees an edge position in consideration of vernier visual acuity. Therefore, in Conventional Example 2, the edge position cannot be guaranteed. Alternatively, in order to forcibly guarantee the edge position, the quantization step width must be made sufficiently small to reduce the compression ratio.

The present invention has been made to solve the above-mentioned problem, and guarantees a perceptible resolution, that is, an edge position defined by vernier visual acuity, and also has a perceptible number of gradations. It is an object of the present invention to provide an image coding technique that can guarantee both of them and that can perform coding with high efficiency.

[0028]

Hereinafter, the present invention will be described. First, the general concept of the present invention will be described.

According to the present invention, input image data is divided into first image data (image data of a small number of gradations representing a perceptible edge position) and second image data (a small number of gradations representing a perceptible gradation number). (Low-resolution image data).

First, the first image data (image data of a small number of gradations representing the perceivable edge position) will be described.

In order to express all perceivable edge positions, the error of the edge position when the gradation level of the input pixel is converted into the edge position may be set within the threshold value of the vernier visual acuity. The edge position error serving as the vernier visual acuity threshold can be converted to a gradation level. By quantizing the input pixels at this gradation level, it is possible to reduce the data amount while guaranteeing the edge position. As a method of obtaining data with the smallest possible number of gradations that can represent all perceived edge positions, quantization of input pixels is performed.

For example, assume that the resolution of the input image is f8.
Also, it is assumed that it is known that the accuracy of dividing the pixels of the input image into 16 should be maintained in order to guarantee the edge position so as to be equal to or smaller than the vernier visual acuity threshold value. At this time, the first image data is image data in which the number of pixel levels of the input image is reduced to 4 bits.

Next, the second image data (low-resolution image data representing the number of perceivable gradations) will be described.

In the visual frequency characteristic diagram shown in FIG. 13, the horizontal axis represents frequency (cycle / degree), and the vertical axis represents contrast sensitivity. The contrast sensitivity indicates the number of perceivable gradations. Here, the number of bits B required to express the perceivable number of gradations N is

## EQU2 ## where B = log ₂ N. The maximum value of the pixel period T that can represent the frequency W is given by the sampling theorem.

## EQU3 ## T = 1 / (2W). The image resolution per 1 degree of viewing angle is represented by f (pixels /
degree), f is

## EQU4 ## f = 1 / T = 2W. From the visual frequency characteristic graph shown in FIG.
The relationship between the image resolution f (number of pixels / degree) and the number of bits B required to express the perceivable number of gradations N can be determined. FIG. 8 shows a conceptual diagram of the relationship between f and B. In the graph of FIG. 8, portions having a low resolution are omitted. If only the high resolution part is taken, the required number of bits decreases as the resolution increases.

By using FIG. 8, it is possible to obtain image data having a resolution as small as possible and capable of expressing all perceivable gradation numbers. In FIG. 8, which is a conceptual diagram, (1) it is sufficient to have 1-bit information when the image resolution is f8; (2) it is sufficient to have 2-bit information when the image resolution is f7 to f8. (3) In the case of image resolutions f6-f7, it is sufficient to have 3-bit information; (4) In the case of image resolutions f5-f6, it is sufficient to have 4-bit information; (5) Image resolution In the case of f4 to f5, it is sufficient to have 5-bit information; (6) In the case of image resolutions f3 to f4, it is sufficient to have 6-bit information; (7) In the case of image resolutions f2 to f3 (8) In the case of image resolutions f1 and f2, it is sufficient to have 8-bit information. Data compression can be achieved by dividing an input image into images of a plurality of resolutions and performing quantization suitable for each resolution. Since the information for each resolution is not independent but includes shared information, it is possible to further compress by having information so that the shared information is omitted as much as possible. In addition, since images of each resolution have a degree of redundancy, further compression is possible by applying a conventional image coding method or data compression method.

Next, overlapping portions between the first image data and the second image data are reduced. According to the first image data,
If a part of the second image data can be represented, the second
The image data is omitted for that part. For example, if the input resolution is f8 and the number of bits after the first image data quantization is 4
In the case of (2), information having an image data amount of 4 bits or less can be omitted from the second image data. That is, (1) 1-bit information when the image resolution is f8, (2) 2-bit information when the image resolution is f7 to f8, (3) 3-bit information when the image resolution is f6 to f7. In the case of resolutions f5 to f6, four pieces of 4-bit information are the first image data, that is, the image resolution is f
8 can be represented by 4-bit information, and is omitted.

Therefore, an image can be expressed with the following information without image quality distortion. (1) Information of image resolution f8, 4 bits (2) Information of image resolution f4 to f5, 5 bits (3) Information of image resolution f3 to f4, 6 bits (4) Information of image resolution f2 to f3, 7 bits Information (5) Image resolution f1 to f2, 8-bit information

For example, "image resolution f4-f5, 5-bit information" may have the highest resolution of this information. That is, it can be represented by “image resolution f5, 5-bit information”. Other information can be summarized as follows: (1) Image resolution is f8, 4-bit information (2) Image resolution f5, 5-bit information (3) Image resolution f4, 6-bit information (4) Image resolution f3, 7-bit (5) Image resolution f2, 8-bit information.

Further, for simplicity, the image resolution f
5 to the image resolution f2,
Also, it is possible to use 8-bit information. In this case, (1) first image data, image resolution is f8, 4-bit information; and (2) second image data, image resolution is f5, 8-bit information.

The concept of the present invention has been described above. Hereinafter, means of the present invention will be described in accordance with each claim.

An image coding apparatus according to claim 1 includes: a pixel value level number reducing unit for reducing the pixel value level number of an input image;
Image reducing means for reducing an input image; first coding means for coding an image whose pixel value level has been reduced by the pixel value level number reducing means; coding of an image reduced by the image reducing means And second encoding means.

In the image encoding apparatus according to the first aspect, an input image is divided into two codes. Using the above example, for example, a first code, an image resolution of f8, encoding a 4-bit image, and a second code, an image resolution of f5, encoding an 8-bit image, The two.

Here, the operation of reducing the image has the same value as the operation of reducing the image resolution.

According to a second aspect of the present invention, there is provided an image coding apparatus comprising: first code amount measuring means for measuring a code amount of a code encoded by a first encoding means; and measurement by the first code amount measuring means. First coding parameter determining means for determining a coding parameter according to the code amount thus obtained; wherein the first coding parameter determining means is measured by the first code amount measuring means. If the code amount is equal to or less than the desired code amount, the coding is performed with the coding parameter unchanged,
When the code amount measured by the first code amount measuring means is larger than a desired code amount, it is determined that the coding parameter is changed and coding is performed again.

An image coding apparatus according to a second aspect controls the code amount of the first coding means.

According to a third aspect of the present invention, there is provided an image coding apparatus comprising: a second code amount measuring means for measuring a code amount encoded by the second coding means; and a second code amount measuring means for measuring the code amount. Second to determine coding parameters according to code amount
Encoding parameter determination means;
When the code amount measured by the second code amount measuring unit is equal to or smaller than a desired code amount, the encoding parameter determining unit encodes the encoded parameter as it is, and performs the second code amount measurement. When the code amount measured by the means is larger than the desired code amount, it is characterized in that it is determined that the coding parameter is changed and coding is performed again.

According to a third aspect of the present invention, there is provided an image coding apparatus for controlling a code amount of the second coding means.

According to a fourth aspect of the present invention, the image coding apparatus is characterized in that the code amount is measured for each line of the input image. The code amount is controlled for each line by measuring the code amount for each line.

According to a fifth aspect of the present invention, the image coding apparatus is characterized in that the code amount is measured for each line of the reduced image. The code amount is controlled for each line of the reduced image by measuring the code amount for each line of the reduced image.

According to a sixth aspect of the present invention, there is provided an image coding apparatus comprising: a first image obtained by reducing the number of pixel value levels for a predetermined image;
A second image obtained by reducing the predetermined image is encoded. Of course, information that is duplicated in encoding the first image and the second image can be reduced.

An image decoding apparatus according to claim 7 is:
A first decoding unit that decodes a first code output from the first image encoding unit of any one of the image encoding devices described above;
Pixel value level number restoring means for converting the pixel value level number of the image decoded by the first decoding means into the level number of the decoded image; output from the second image encoding means of the image encoding device Second decoding means for decoding the second code; image enlargement means for converting the size of the image decoded by the second decoding means into the image size of the decoded image;
At least a decoded image generating means for inputting the pixel value of the image restored by the pixel value level number restoring means and the pixel value of the image restored by the image enlarging means and generating a decoded image is provided. It is characterized by.

An image decoding apparatus according to claim 8, further comprising: image reducing means for reducing the decoded image output from the decoded image generating means to have the same resolution as the image decoded by the second decoding means. An image input unit for inputting an image decoded by the second decoding unit; and a difference between a pixel value of the image reduced by the image reduction unit and a pixel value of the image input by the image input unit. Difference calculating means for calculating; adding value calculating means for calculating an added value to be added to each pixel value from the difference value calculated by the difference calculating means; and adding a pixel to the decoded image output from the decoded image generating means. And a pixel value adding means for adding the addition value calculated by the addition value calculation means to the value.

The present invention can be implemented not only as an apparatus or a system but also as a method.
At least a part thereof can be implemented as a computer program.

The above features of the present invention and other features of the present invention are set forth in the appended claims, and will be described in detail below with reference to embodiments.

[0055]

DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, the present invention will be described in detail with reference to embodiments.

[Embodiment 1] Embodiment 1 relates to an encoding apparatus without a code amount control function (FIG. 1). Also,
The present invention relates to a decoding device (FIG. 2) for decoding an image encoded by the encoding device.

Hereinafter, the data width per pixel of the input image is 8 bits, and the resolution is 600 dpi (hereinafter, 25.4 m).
An image resolution in which N pixels are arranged on a straight line of m is described as Ndpi). The distance between pixels of 600 dpi is about 4
2.3μ. Assuming that the visual distance is 30 cm, the distance at which the threshold value of the vernier visual acuity is 2 ″ is 30 sin (2 ″) cm. This is calculated to be about 2.9 μ. 42.3
Since /2.9=approximately 15, the pixel value of 600 dpi can be controlled to be equal to or less than the threshold value of the vernier visual acuity if it can express approximately 15 or more gradations. Therefore, 600 dpi is represented by 4 bits.

Further, the vertical axis of the graph of FIG.
pi). When the viewing distance is set to 30 cm and deformed, for example, the graph can be rewritten as shown in FIG. In FIG. 9, 200 dp
It suffices that 8 bits be used to represent an image of a frequency that can be represented with a resolution of i or less.

An image of a frequency that can be expressed with a resolution from 200 dpi to 600 dpi can be expressed by expressing 600 dpi with 4 bits.

That is, in view of the relationship between the viewing distance and the visual frequency characteristic, if the characteristic shown in FIG.
In other words, the 200 dpi image may be represented by 8 bits. In the first embodiment, a specific example for realizing the two-division image encoding method will be described.

Hereinafter, the first embodiment will be described in detail with reference to FIG.

In FIG. 1, reference numeral 101 denotes an input image;
Is a quantization LUT (lookup table) for quantizing an input image, 103 is a first encoder for encoding output data of the quantization LUT 102, 104 is a first code, 105
Is a reduction circuit for reducing an input image, 106 is a reduction circuit 10
A second encoder 107 for encoding the output data of No. 5 is a second code.

Hereinafter, the operation of the encoding apparatus according to the first embodiment will be described.

An input image 101 having a resolution of 600 dpi is input to a quantization LUT 102 and a reduction circuit 105.

FIG. 6 shows an example of the quantization LUT 103. 8-bit data is input, and 4-bit data is output. Here, the LUT is set so that when the input is 0, the decoded image is 0, and when the input is 255, the decoded image is 255. That is, the index is 0 only when the input value is 0, and the index is 15 only when the input value is 255.

The reduction circuit 102 converts the input image 101 into
The size is reduced to one third in each of the vertical and horizontal directions. For example, the following method is used.

The image input in raster order is divided into blocks of 3 × 3 pixels. A 600 dpi pixel block of 3 × 3 pixels corresponds to one pixel having a resolution of 200 dpi. By adding all the pixel values in the 3 × 3 block and dividing the added value by 9, it is possible to obtain a reduced image obtained by reducing the input image 101 vertically and horizontally by one third. The above is one example of obtaining a reduced image. By applying a general low-pass filter to the input image 101 and sub-sampling the output, a higher-quality reduced image with small aliasing distortion can be obtained.

The quantization index quantized by the quantization LUT 102 is encoded by the first encoder 103 to become the first code 104. As the encoding method in the first encoder 103, a lossless encoding method such as the LZ method is suitable. A binary reversible method such as the JBIG method, the MH, and the MMR method may be applied by decomposing the image into bit-branes. Or, JP
It is also possible to apply an irreversible image encoding method such as the EG method.

The reduced image reduced by the reduction circuit 105 is encoded by the second encoder 105 to become a second code 107. As the encoding method in the second encoder 106, for example, a lossy image encoding method such as the JPEG method can be adopted. Alternatively, it is also possible to adopt a lossless encoding method such as the LZ method. A binary reversible method such as the JBIG method, the MH, and the MMR method may be applied by decomposing the image into bit-branes.

The first code 104 and the second code 107 are
It may be multiplexed into one code stream.

The operation of the encoding apparatus according to the first embodiment has been described above.

Next, the operation of the decoding apparatus according to the first embodiment will be described.

FIG. 2 is a diagram showing the configuration of the decoding apparatus according to the first embodiment.

In FIG. 2, reference numeral 201 denotes a first code,
2 is a first decoder, 203 is an inverse quantization LUT, 204 is a second code, 205 is a second decoder, 206 is an enlargement circuit, 2
07 is a selection circuit, 208 is a decoded image, 209 is a dequantized image, and 210 is an enlarged image.

Hereinafter, the configuration and operation of the decoding apparatus according to the first embodiment will be described with reference to FIG.

The first code 201 has the same format as the first code 104 encoded by the first encoder 103 in FIG. The second code 204 has the same format as the second code 107 encoded by the second encoder 106 in FIG.

The first code 201 is decoded by the first decoder 202, and a quantized index is output. The quantization index is inversely quantized by the inverse quantization LUT 203, and an inversely quantized image 209 is obtained. FIG. 7 shows an example of the contents of the inverse quantization LUT. The content of the inverse quantization LUT 203 is the quantization LU
It must match the contents of T102.

The second code 204 is the second decoder 2
05, and a reduced image is output. The reduced image is enlarged by the enlargement circuit 206, and an enlarged image 210 is output. Various methods such as the nearest neighbor method, the bilinear method, and the cubic convolution method can be applied as the enlargement method.

The inversely quantized image 209 and the enlarged image 210 are input to the selection circuit 207.

In the selection circuit 207, the inverse quantized image 209
And an expanded image 210 to generate and output a decoded image 208. Hereinafter, an operation example of the selection circuit 207 will be described.

The pixel value of the inversely quantized image is P, and the pixel value of the enlarged image is L. The predetermined threshold value is T, the output value of the selection circuit 207 is Y, and the function for calculating the absolute value is abs ().

[Mathematical formula-see original document] When abs (PL) ≤T, Y = P when abs (PL)> T.

Another example will be described. It is assumed that predetermined thresholds T1 and T2 (where T2 ≧ T1 ≧ 0).

When abs (P−L) ≦ T1, Y = L T1 <P−L ≦ T2, Y = P−T1 When T1 <LP ≦ T2, Y = P + T1 abs (P− L)> T2, Y = P

The above threshold value T, or T1, T2, may be a function of the quantization index output from the first decoder. This function can be realized by an LUT.

Alternatively, it is also possible to design an LUT for directly obtaining the output value Y by inputting the difference value PL and the quantization index. In this case, since the threshold processing is not performed as described above, the output value Y can be freely designed.

The description of the operation of the first embodiment has been completed.

As described above, in the first embodiment, it is possible to control the number of gradations of the pixel of 600 dpi which is the resolution of the input image to a value that guarantees the edge position defined by the vernier visual acuity. In addition, it is possible to guarantee a desired (for example, 8 bits) gradation number with an image resolution of 200 dpi. The 600 dpi pixel value can be reduced from 8 bits to 4 bits. Also, a pixel value of 200 dpi is 8 bits, and when converted to 600 dpi, 8 pixels per pixel.
/ 9 bits. That is, it is possible to perform image compression from 8 bits to 4 + 8/9 bits without performing special image compression. Further compression can be performed, in which case 4 + 8/9 bits can be further reduced.

In the first embodiment, as shown in FIG. 9, the case where a 200-dpi frequency is required to be 4 bits happens to occur. It is necessary to set a sufficient number of gradation bits from among the combinations. As shown in FIG. 10, even when four bits are not required at 200 dpi, it is necessary to use a 4-bit code at 200 dpi as an encoding method.

In the above example, the non-linear quantization is performed using the LUT. However, it is needless to say that linear quantization using ordinary division may be used.

Alternatively, the upper bits may be directly extracted and used as a quantization index. In this case, the effect of reducing the calculation load can be obtained.

Alternatively, a special quantization index is given only to a special value (for example, 0 and 255) as the input value.
In other cases, a method of extracting the upper bits as they are and using them as quantization indexes may be used.

In the first embodiment, the quantization LUT 102
Is used as the quantization index, and the first code is the code of the quantization index. However, the output of the quantization LUT may be a quantization representative value instead of the quantization index. In this case, the pixel value of the quantized representative value has a higher probability of the same value continuing than the pixel value of the input image 101, so that the compression ratio such as run length can be increased. Therefore, even when the quantization representative value is adopted, high-efficiency compression can be performed similarly to the case where the quantization index is adopted. When the quantization representative value is encoded, there is an effect that the inverse quantization LUT of the decoder becomes unnecessary.

[Embodiment 2] The coding apparatus according to the second embodiment is obtained by adding a code amount control function of the first code to the coding apparatus according to the first embodiment.

The configuration and operation of the second embodiment will be described below with reference to FIG. In FIG. 3, reference numeral 301 denotes an image buffer;
02 is a code buffer. 3, elements to which the same reference numerals are added as in FIG. 1 perform the same operations.

The input image 101 is input to the image buffer 301. After the image is accumulated in the image buffer 301 for one line, the image is encoded using the quantization LUT 102 and the first encoder 103, and the code is encoded in the code buffer 302.
Is input to

(1) The code amount input to the code buffer 302 is evaluated.

(2) When the code amount input to the code buffer 302 is equal to or smaller than a predetermined value, the code buffer 30
2 is output as the first code 104.
Along with the first code 104, an encoding parameter used when the code is generated is also added to the code. This operation allows decoding at the decoder. When the code amount input to the code buffer 302 is equal to or more than a predetermined value, the code stored in the code buffer 302 is discarded, and the image buffer 3
01 is re-encoded. When encoding is performed again, the encoding parameters are changed. For example, the compression ratio can be increased by changing the number of pixel value levels. In this case, the contents of the quantization LUT shown in FIG. 6 are changed to reduce the maximum value of the quantization index.

(3) After the coding is completed, the code buffer 302
Is re-evaluated. Return to (2).

The above steps (2) and (3) are repeated to control so that the code amount is equal to or less than a predetermined code amount.

On the other hand, the input image 101 is reduced by the reduction circuit 105.
Is input to The subsequent operation is the same as in the first embodiment.

The decoder decodes the quantization index of each line and the coding parameter of the line, and decodes each line using the coding parameter.

In the second embodiment, the code amount control is performed for each line of the input image. However, the unit for performing the code amount control may be anything other than the line, such as a band in which a plurality of lines are combined, a block in which the image is blocked, and the like. good.

[Embodiment 3] The coding apparatus according to the third embodiment is obtained by adding a code amount control function of the second code to the coding apparatus according to the first embodiment.

The configuration and operation of the third embodiment will be described below with reference to FIG. In FIG. 4, reference numeral 401 denotes an image buffer;
02 is a code buffer. In FIG. 4, elements to which the same reference numerals are added as in FIG. 1 perform the same operations.

The input image 101 is input to the image buffer 401. First, it is assumed that reduction is performed in 2 × 2 blocks. After the image is stored in the image buffer 301 for two lines, the image is encoded using the reduction circuit 105 and the second encoder 106, and the code is input to the code buffer 402.

(1) The code amount input to the code buffer 402 is evaluated.

(2) When the code amount input to the code buffer 402 is equal to or less than a predetermined value, the code buffer 40
2 is output as the second code 107.
Along with the second code 107, an encoding parameter used when the code is generated is also added to the code. This operation allows decoding at the decoder. When the code amount input to the code buffer 402 is equal to or larger than a predetermined value, the code stored in the code buffer 402 is discarded, and the image buffer 4
01 is newly added to the input image, and re-encoded together with the already stored image. When encoding is performed again, the encoding parameters are changed. For example, the compression ratio can be increased by changing the reduction ratio. In this case, one or more lines are further input to the image buffer 401 so that a desired reduction ratio can be obtained. When changing the reduction ratio from 1 / N to 1 / M (M> N), the MN line input image is input to the image buffer.

(3) After the encoding is completed, the code buffer 302
Is re-evaluated. Return to (2).

The above steps (2) and (3) are repeated to control the code amount to be equal to or less than the predetermined code amount.

On the other hand, the input image 101 is a quantized LUT 1
02 is input. The subsequent operation is the same as in the first embodiment.

The decoder decodes the pixel value of each reduced line and the coding parameters of the reduced line, and decodes each line using the coding parameters.

In the third embodiment, the code amount control is performed for each line of the reduced image. However, the unit for performing the code amount control may be a band obtained by combining a plurality of lines of the reduced image.

In the second and third embodiments, only one of the code amount of the first code and the code amount of the second code is controlled. However, a configuration in which both are controlled may be employed.

[Fourth Embodiment] In the fourth embodiment, a circuit for storing an average value for each block is added to the image decoder shown in the first embodiment.

The configuration and operation of the image decoding apparatus according to the fourth embodiment will be described with reference to FIG. 5, the same components as those of FIG. 2 are denoted by the same reference numerals. The newly shown components are: Reference numeral 501 denotes an output value of the selection circuit 207. This is the same as the decoded image in the first embodiment. A reduction circuit 502 reduces the output value of the selection circuit 207 to create a reduced pixel 508, and a reference numeral 503 denotes a second decoder 2
The added value calculation circuit 504 receives the decoded image 507 of FIG. 5 and the output value of the reduction circuit to calculate the added value, 504 is the added value calculated by the added value calculation circuit 503, and 505 is the selection circuit 2
An addition circuit 504 adds the addition value 504 to the output value 501 of 07, and 505 is a decoded image.

The output value 507 of the second decoder 205 is a reduced image of the original image. This reduced image is generated, for example, by dividing the original image into 3 × 3 blocks and calculating the average of the pixel values in the blocks (described in Example 1). This image 507 is obtained by adding the
3 is input.

The output value 501 of the selection circuit 207 is input to the reduction circuit 502, and a reduced image 508 is created. The reduction method in the reduction circuit 502 performs the same operation as the reduction circuit 105 in the image decoding device according to the first embodiment. The reduced image 508 is input to the addition value calculation circuit 503.

The reduction operation is performed by a method of calculating the average value of the blocks (described in the first embodiment). The addition value calculation and addition are performed for each block. Hereinafter, the number of pixels in a block is set to N. It is assumed that the pixel value of the image 507 and the pixel value of the reduced image 508 existing in a certain block are Ai (where 1 ≦ i ≦ N) and Bi (where 1 ≦ i ≦ N). In the addition value calculation circuit 503,

(Equation 7) By the calculation of, an addition value 504 (s in the above equation) is calculated. The addition value 504 is calculated by the addition circuit 505.

[Mathematical formula-see original document] The operation of Bi '= Bi + s is performed to obtain the final decoded image Bi'. Bi ′ becomes the decoded image 506.

[0118]

As described above, according to the image encoding method of the present invention, the input image data is combined with the first image data (the image data of a small number of gradations representing the perceptible edge position). And the second image data (low-resolution image data representing the perceivable number of tones) and encoding the data so that the resolution and the number of tones required to represent the image are reduced to as small as possible. Can be realized in quantity. Further, even when the gradation level represents the edge position within one pixel, the edge position is held by holding the image data of the smallest possible number of gradations capable of expressing all the perceivable edge positions. This has the effect that it can be guaranteed.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating a configuration of an image encoding device according to a first embodiment of the present invention.

FIG. 2 is a block diagram illustrating a configuration of an image decoding device according to the first embodiment.

FIG. 3 is a block diagram illustrating a configuration of an image encoding device according to a second embodiment of the present invention.

FIG. 4 is a block diagram illustrating a configuration of an image encoding device according to a third embodiment of the present invention.

FIG. 5 is a block diagram illustrating a configuration of an image decoding device according to a fourth embodiment of the present invention.

FIG. 6 is a diagram illustrating an example of a quantization LUT according to the above embodiment.

FIG. 7 is a diagram illustrating an example of an inverse quantization LUT according to the above embodiment.

FIG. 8 is a conceptual diagram illustrating the relationship between the required number of bits and the image resolution according to the present invention.

FIG. 9 is a conceptual diagram illustrating a relationship between a required number of bits and an image resolution according to the first embodiment of the present invention.

FIG. 10 is another conceptual diagram illustrating the relationship between the required number of bits and the image resolution according to the first embodiment of the present invention.

FIG. 11 is a view for explaining Conventional Example 1.

FIG. 12 is a diagram illustrating a second conventional example.

FIG. 13 is a diagram illustrating visual frequency characteristics.

FIG. 14 is a diagram for explaining a character / line drawing and a photograph mixed block.

FIG. 15 is a diagram illustrating a binary font.

FIG. 16 is a diagram illustrating a gray font.

FIG. 17 is a diagram illustrating vernier visual acuity.

[Explanation of symbols]

 Reference Signs List 101 input image 102 quantization LUT 103 first encoder 104 first code 105 reduction circuit 106 second encoder 107 second code 201 first code 202 first decoder 203 inverse quantization LUT 204 second Code 205 second decoder 206 enlargement circuit 207 selected image 208 decoded image 209 inverse quantized image 210 enlarged image 301 image buffer 302 code buffer 401 image buffer 402 code buffer 501 output value of selection circuit 207 502 reduction circuit 503 addition value calculation Circuit 504 Addition value 505 Addition circuit 505 Decoded image 507 Decoded image of second decoder 508 Reduced image

 ──────────────────────────────────────────────────続 き Continuing on the front page (72) Inventor Song Kazunori, 430 Nakai-cho, Nakai-machi, Ashigara-gun, Kanagawa Prefecture Inside Fuji Xerox Co., Ltd. F term (reference) 5C059 KK02 KK27 MD07 TA12 TA36 TA53 TB09 TC18 UA02 5J064 AA02 BA15 BB01 BC02 BC08 BC16 BD01

Claims (10)

[Claims] A pixel value level reducing unit configured to reduce the number of pixel value levels of an input image; an image reducing unit configured to reduce the input image; and an image whose pixel value level is reduced by the pixel value level reducing unit. An image encoding apparatus comprising: a first encoding unit that encodes an image; and a second encoding unit that encodes an image reduced by the image reduction unit. 2. A first code amount measuring means for measuring a code amount of a code encoded by the first encoding means, and a code amount measured by the first code amount measuring means. A first coding parameter determining unit for determining a coding parameter, wherein the first coding parameter determining unit determines that the code amount measured by the first code amount measuring unit is a desired code amount. In the following case, the encoding parameter is encoded as it is, and when the encoding amount measured by the first encoding amount measuring means is larger than a desired encoding amount, the encoding parameter is changed. 2. The image encoding apparatus according to claim 1, wherein it is determined that encoding is performed again. 3. A second code amount measuring means for measuring the code amount encoded by the second encoding means, and encoding according to the code amount measured by the second code amount measuring means. A second encoding parameter determining unit that determines a parameter, wherein the second encoding parameter determining unit determines that the code amount measured by the second code amount measuring unit is equal to or less than a desired code amount. In some cases, the encoding parameter is encoded as it is, and when the encoding amount measured by the second encoding amount measuring means is larger than the desired encoding amount, the encoding parameter is changed and the encoding is performed again. The image coding apparatus according to claim 1, wherein the image coding apparatus determines to perform coding. 4. An apparatus according to claim 2, wherein said first code amount measuring means measures the code amount for each line of the input image. 5. An image encoding apparatus according to claim 3, wherein said second code amount measuring means measures the code amount for each line of the reduced image. 6. An image, wherein a first image obtained by reducing the number of pixel value levels for a predetermined image and a second image obtained by reducing the predetermined image are encoded. Encoding device. 7. A first decoding unit for decoding a first code output from a first image encoding unit of the image encoding device according to any one of claims 1 to 5, and the first decoding unit. Pixel value level number restoring means for converting the pixel value level number of the image decoded by the above into a decoded image level number, and decoding the second code output from the second image encoding means of the image encoding device A second decoding unit that converts the size of the image decoded by the second decoding unit into an image size of the decoded image; An image decoding apparatus comprising: a decoded image generating unit that receives a pixel value and a pixel value of an image restored by the image enlarging unit and generates a decoded image. 8. An image reduction means for reducing a decoded image output from the decoded image generation means to have the same resolution as an image decoded by the second decoding means, and Image input means for inputting the decoded image; difference calculating means for calculating a difference between a pixel value of the image reduced by the image reducing means and a pixel value of the image input by the image input means; An addition value calculation unit that calculates an addition value to be added to each pixel value from the difference value calculated by the difference calculation unit; and 8. The image decoding apparatus according to claim 7, further comprising: a pixel value adding unit that adds the addition value calculated by the calculation unit. 9. An image whose pixel value level has been reduced by the pixel value level number reducing step of reducing the number of pixel value levels of the input image, the image reducing step of reducing the input image, and the pixel value level number reducing step. And a second encoding step of encoding the image reduced by the image reducing step. 10. A first decoding step for decoding a first code output in a first image encoding step of the image encoding method according to claim 9, and an image decoded by said first decoding step. A pixel value level number restoring step of converting the pixel value level number into a decoded image level number, and a second decoding for decoding a second code output by the second image encoding step of the image encoding method. An image enlarging step of converting the size of the image decoded by the second decoding step into an image size of the decoded image; and at least a pixel value of the image restored by the pixel value level number restoring step; A decoded image generating step of inputting the pixel value of the image restored by the image enlarging step and generating a decoded image. .