JP3205010B2 - Image compression device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image compression apparatus, and more particularly to an image compression apparatus which combines image information having a gradation such as a natural image and a photograph with artificial character and line drawing information generated by a computer or the like. The present invention relates to an image compression device that performs compression using orthogonal transform.

[0002]

2. Description of the Related Art The memory capacity required to store a halftone image such as a photograph in a memory is (number of pixels) × (number of gradation bits), which is enormous for storing a high-quality color image. Memory capacity was required. For this reason, various information amount compression methods have been proposed, and the amount of information is compressed and then stored in a memory to reduce the memory capacity. In recent years, as an international standardized method of color still image coding, there is a coding method of a Base Line System (basic method) proposed by JPEG (Joint Photographic Experts Group).

[0003] This system is based on human visual characteristics and DCT.
In consideration of the occurrence of (discrete cosine transform) components, the quantization step is set to be small in the low frequency band of the DCT space. Therefore, it is possible to provide an image in which the image quality is not significantly deteriorated even at a high compression ratio for a halftone image.

[0004]

However, the above-mentioned conventional example has the following problems.

That is, when the above-mentioned conventional example is applied to, for example, an image output device, the image output device is generally connected to an image input device such as a host computer or an image scanner and often operates as a system. In this case, various images such as CG (computer graphic) images created on the host computer and images input by the image scanner are transmitted to the image output device.

In the above-mentioned conventional example, as described above, the present invention is effective for an image in which the conversion coefficients are concentrated in the low frequency band of the DCT, such as an image obtained by digitizing an image such as a photograph by an image scanner. In the case of images, font information such as characters, and line drawings such as CAD (Computer Aided Design), large transform coefficients also occur in the DCT high band, and the image expanded by coarse quantization of the high band coefficients is a so-called image. The result is a poor image with mosquito noise ringing.

Therefore, even when an image having different properties as described above is transmitted, we automatically identify a natural image and an artificially generated character, line drawing, etc., and separate them to obtain a character,
For line drawings and the like, the information is stored in a bitmap memory having no depth information, density information (color information) such as characters and line drawings is stored in a gradation memory, and the remaining natural image is subjected to normal DCT and image memory after quantization. We proposed the method of storing in.
According to this method, artificially generated character and line drawing information can be combined after natural image expansion without distortion. In addition, by replacing the pixel where the character and the line drawing existed after separating the character and the line drawing with the average value of the non-separable region, and performing DCT and quantization, the non-separable region (part of the natural image) is also obtained. A good image with little deterioration can be reproduced.

However, in the above proposal, automatic identification and separation of artificially generated characters, line drawings, and the like in a natural image has not been proposed in a method that is simple and has few erroneous detections.

SUMMARY OF THE INVENTION The present invention has been made to solve the above-mentioned problem, and has an extremely simple structure and can accurately identify and separate pixels constituting a character or a line drawing without creating a histogram or the like. It is an object to provide a compression device.

[0010]

In order to achieve the above object, the present invention provides a method for compressing an image in which a natural image and a line image are mixed.
An image compression apparatus for compressing input image data to a predetermined size.
Blocking means for blocking into blocks;
Based on each block obtained by the
Extraction to extract image data having prime values as line images
Means, and pixel values of the image data extracted by the extraction means
Is replaced with another pixel value.
Compression means, and the extraction means includes
Large value (MAX1), minimum value (MIN1),
The second largest value (MAX2), the second smallest value
Extract the four values (value MIN2)
Calculate MAX2, MIN2-MIN1, and calculate the subtraction value.
Depending on the result of comparison with each other and a predetermined threshold, MAX1
Or to determine which of MIN1 to extract.
Sign . Further, in order to achieve the above object, the present invention
Is an image that compresses an image in which natural images and line images are mixed
A compression device for converting input image data into predetermined blocks.
Blocking means for blocking, and
Has a predetermined pixel value based on each block obtained
Extracting means for extracting image data as a line image,
The pixel values of the image data extracted by the extraction means are
Compression means for compressing the image data after replacement with prime values
The extraction means has a maximum value (MA) in the block.
X1), minimum value (MIN1), second largest
(MAX2), the second smallest value (MIN
2) are extracted, and (MAX1-MAX)
2) / (MAX2-MIN1), (MIN2-MIN)
1) Calculate / (MAX1-MIN2), and calculate
MAX1 or less depending on the result of comparison with the predetermined threshold value.
Or MIN1 is determined.
And

[0011]

[0012]

[0013]

[0014]

[0015]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS A preferred embodiment according to the present invention will be described below in detail with reference to the drawings.

FIG. 1 is a schematic block diagram showing the configuration of an image compression apparatus according to this embodiment. In the figure, reference numeral 100 denotes an input terminal to which a multi-level image signal is input. Here, the input signal is image data in which a halftone image having a gradation, a CG image, a character, a line drawing, and the like created by a computer are combined without distinction. The input multi-level image signal is stored in a line buffer (not shown) for eight lines, and is cut out by the blocking circuit 101 into a block of 8 × 8 pixels. In the present embodiment, an explanation will be given using an 8 × 8 DCT for the orthogonal transform, but it is a matter of course that other orthogonal transform schemes or block sizes other than 8 × 8 may be used.

Reference numeral 102 denotes an extraction circuit in the block, and 1
The highest level (MAX1) in the block 64 pixels
The second highest level (MAX2), the lowest level (MIN1), and the second lowest level (MIN2) are extracted. Here, when 64 pixels are at the same level in the block, MAX1 = MAX2 = MIN2 = M
If there are only two levels, then MAX1 =
MIN2, MAX2 = MIN1, and in the case of three levels, MAX1, MAX2 = MIN2, MIN1 are three levels, and in the case of four or more levels, MAX1> MAX
2>MIN2> MIN1 holds. This extraction circuit is composed of four registers MAX1, MAX2, MIN2, and MIN1 and a comparator, and four levels are obtained at the same time when scanning of 64 pixels is completed. Reference numeral 103 denotes an adder, which is a circuit for obtaining an added value for all levels of the input 64 pixels. In the case of the present embodiment, since the gradation of the input pixel is 8 bits and 64 pixels per block, the added result is a value within a maximum of 14 bits.

Numerals 104a and 104b denote counters. Counters A and B respectively count how many pixels MAX1 and MIN1 extracted by the extraction circuit 102 are in one block. And 10
5a and 5b are MAX1 and MI extracted by the extraction circuit 102.
Indicates where N1 is located in the block 6
4 shows a bit map for 4 bits. That is, MAX1
The bitmap 105a sets the pixel corresponding to MAX1 in the block to “1” and sets the other pixels to “0”. The MIN1 bitmap 105b sets the pixel corresponding to MIN1 in the block to “1” and sets the other pixels to “1”. A bitmap is created as "0". The extraction circuit 102 scans 64 pixels.
Registration and update to each register are repeated. When MAX1 is updated, the counter A 104a and the bit map 105 are updated.
a is reset and the counter B10 is updated when MIN1 is updated.
4b and the reset of the bitmap 105b are of course linked.

Reference numeral 106 denotes a judgment circuit, and the extraction circuit 102
Is to separate the pixel part of MAX1 extracted in
Alternatively, the pixel portion of MIN1 is separated, or
5 shows a determination circuit for determining whether to select three types of separation. The details of the determination circuit 106 will be described later. Then, the determination result is transmitted to the switch 115, the switch 107, and the switch 108. That is, the judgment result is MAX
If it is determined that one pixel is separated, the counter A 104 counts the number of MAX1 pixels in the switch 115.
a is selected, and the MAX1 bitmap 105 a is selected by the switch 107 and stored in the bitmap memory 109. Similarly, the judgment result is MIN1
When it is determined that the pixels are separated, a counter B 104b that counts the number of pixels of MIN1 by the switch 115
Is selected, and the MIN bitmap 105b is selected by the switch 107.
9 is stored. If it is determined that the separation is not performed, the bit stuffing 11
0 is selected, and all “0” for 64 pixels are stored in the bitmap memory 109. The above-described bitmap memory 109 stores a pixel portion separated in each block as "1", and stores the (pixel) bits of the input image.

If it is determined that either MAX1 or MIN1 is to be separated, the switch 108
In the figure, the lower side (terminal B) is selected, and information of a block delayed by 64 pixels in the one-block delay circuit 111 (information of the same block as the block used for extraction and determination) is separated and replaced. The signal is input to the circuit 112.
That is, since one block is scanned by the extraction circuit 102, information of the same block is input with a delay of one block (the time in the above determination is ignored).
If it is determined that the determination result is not separated, the terminal A is selected by the switch 108, and the information in one block is transmitted to the DCT circuit 113 without any processing. When one of MAX1 and MIN1 is selected by the determination circuit 106, the selected level value is stored in the gradation memory 114 and saved. Here, the gradation memory 114 is allocated for 8 bits per block and holds (pixels / 64) bytes of the input image.

The separation / replacement circuit 112 is a circuit that separates a pixel corresponding to the address where “1” occurs by reading out the bitmap memory 109 and replaces the pixel with a predetermined value. In this embodiment, a method will be described in which the replacement value is an average value of the levels of the pixels that are not separated out of 64 pixels. That is, the data from the adder 103 input to the separation / substitution circuit 112 is T, and the counter A 104a,
Alternatively, when the input from the counter B 104b is C and the input from the gradation memory 114 is G, the replacement value TR is calculated by the following equation.

TR = (T−C × G) / (64−C) That is, all the separated pixels are replaced with the value of the replacement value TR and input to the DCT circuit 113. And D
The DCT (Discrete Cosine Transform) is performed by the CT circuit 113, output to the output terminal 116, and the DCT coefficients are stored in a memory after quantization (not shown).

After the encoding according to the above configuration is performed,
At the time of decoding, on the other hand, on the signal after IDCT (inverse discrete cosine transform), based on the bitmap memory, the value of the gray scale memory of the block is substituted for the pixel where "1" is generated, The separated portion can be restored without deteriorating both resolution and gradation.

FIG. 2 is a block diagram showing a detailed configuration of the determination circuit 106 described above. As shown in the figure, the input signals MAX1, MAX2, MIN2, and MIN1 are subtracted by subtracters 201 and 202 into (MAX), respectively.
1-MAX2), (MIN2-MIN1) are calculated. Then, the (MAX1-MAX2), (MIN2
−MIN1), the magnitude relation is compared by the comparator 203,
The switch 204 selects the larger value. The selected signal is applied to a comparator 205 at a certain threshold (T
H) and a determination is made as to whether or not separation is to be performed. That is, (MAX1-MAX2) or (MIN2-M
When the difference in contrast of IN1) is large to some extent,
The one with the larger contrast difference is the object of separation.

The above determination will be described by taking actual image information as an example. FIG. 3A shows a certain natural image 8b.
A line drawing is artificially generated from a computer on the it data and a certain block of the synthesized image is extracted. In this block, the outputs from the extraction circuit 102 are MAX1 = 171, MAX2 = 170, M
IN2 = 104 and MIN1 = 20. When this is calculated by the determination circuit of FIG. 2, MAX1-MAX2 =
1, MIN2-MIN1 = 84, and MIN2-MI
N1 is compared with a certain threshold TH. Here, for example, TH
= 40, MIN2-MIN1> T
H holds, and the pixels of MIN1 are separated on the assumption that they are artificially generated pixels.

FIG. 3C shows the separated MIN.
1, that is, the average value “14” of 55 pixels (64−9 = 55) not separated at the position of the pixel where “20” existed.
6 "is a block obtained by substituting and replacing 6". (B) and (d) shown in FIG. 3 correspond to (a) and (c) shown in FIG.
FIG. 4 is a diagram illustrating a state of generation of a DCT component when a CT is applied (11-bit accuracy, the DC component is offset from an input pixel).

As is apparent from a comparison between (b) and (d) shown in FIG. 3, when the separation is successful, a large coefficient in the DCT high band is removed, and a code in which the generation of the coefficient is concentrated in the low band. It can be seen that good quantization with high conversion efficiency can be performed.

Next, (a) and (b) shown in FIG.
Similarly to (a) shown in FIG. 7, a line drawing generated by a computer is synthesized on a natural image at a value of "20". In the case of FIG. 4A, MAX1 = 74, MAX2 = 6.
4, MIN2 = 12, and MIN1 = 10, and it is determined by the determination circuit of FIG. 2 that separation is not performed. In the case of FIG. 4B, MAX1 = 88, MAX2 = 70, M
IN2 = 32 and MIN1 = 20, but TH> (MA
X1−MAX2)> (MIN2−MIN1), and it is determined that separation is not performed as in (a) of FIG. 4 (TH
= 40). However, automatic separation of signals originally generated on a computer originally prevents ringing such as mosquito noise due to coarse quantization of high-frequency components during quantization, and allows visual recognition without increasing the code amount. The purpose is to prevent a typical deterioration in image quality. That is, in the blocks (a) and (b) shown in FIG. 4, even if the value "20" cannot be separated, there is no significant effect for the following reason.

First, the input image itself is "20".
Line drawing information is inconspicuous. That is, in the case of (a) shown in FIG. 4, there are pixels having a level of “20” or more in one block, and pixels having a level of “20” or less in one block. It will exist across levels. Further, in the case of (b) shown in FIG. 4, although the level does not cross the level “20”, there is no contrast difference, and the natural image is formed by the neighboring levels. Since the original line drawing of “20” is inconspicuous, even if the AC component is quantized by the quantization after the DCT, the amplitude is slightly changed but does not have a great effect visually.

The second reason is that there is not much change in the difference in contrast due to the combination of the line drawing “20” (it does not become a large contrast edge).
This is because even if DCT is performed without separation, the high-frequency AC component does not occur with a large coefficient. The cause of the image quality deterioration is mainly due to the quantization error of the high frequency AC component.
Since the C component is small, the influence of the quantization error can be small.
That is, DCT, quantization, inverse quantization,
Even if the IDCT is performed, the visual deterioration of the output image is small.

FIG. 5 shows a histogram of the above-described determination for easy understanding. The case where the histogram of one block is divided into two as shown in FIG. 5, that is, the case where a cluster of natural image levels and an artificial line drawing of the same level are separated, is defined as MAX1, M2.
AX2, MIN2, and MIN1 are extracted and determined. In general, when DTP (desktop publishing) is considered, it is often the case that a character and a line image to be synthesized in a natural image use a different level from the levels of the natural image and the background, that is, a value equal to or more than a certain level difference. Seem.

As described above, according to this embodiment,
It is possible to realize automatic extraction and separation of characters, line drawings, and the like that are artificially generated by an extremely simple method without creating a histogram of levels in a block. <Modification 1 of Judgment Circuit> FIG. 6 is a block diagram showing a configuration of a judgment circuit 106 in a modification. In this example,
This is a case where the outputs from the counters A, B 104a, b shown in FIG. In FIG. 6, MAX1count is an input from the counter A 104a, and MIN1count is an input from the counter B 104b. MAX1, MAX2, MIN2,
The input of MIN1 is the same as in FIG. The same parts as those in FIG. 2 are denoted by the same reference numerals.

In the circuit of FIG. 6, (MAX1-MAX2),
(MIN2-MIN1), and the magnitude relation is calculated by the comparator 20.
3 is the same as that of the circuit of FIG. 2, except that the comparison result is transmitted to the switch 601. That is, in the switch 601, (MAX1-
When MAX2) is selected, MAX1count is selected. When (MIN2-MIN1) is selected, MIN1count is selected.
Select count. Switch 204 and comparator 20
5 is the same as the circuit of FIG. 2, and the description is omitted, but MAX1 or MIN1 selected by the switch 601 is omitted.
The number of pixels in the block is determined by a comparator 602 at a certain threshold (TH
2) is compared.

That is, in this determination circuit, the extracted M
The condition that AX1 or MIN1 must be present in the number of (TH2) pixels or more in one block is added. For example, even if the condition shown in FIG. If there is a pixel having a value that is only an extraordinary pixel, it is possible to determine whether the pixel is an artificially generated character or line drawing or is due to noise in the input image. This depends on the characteristics of the system equipped with the image compression device of the present invention. In this example,
AN is determined by the two conditions of the comparator 205 and the comparator 602.
A determination is made through the D-circuit 602 as to whether or not to separate. <Modification 2 of Judgment Circuit> FIG.
It is a block diagram which shows the modification of FIG. As shown
The extracted MAX1, MAX2, MIN2, MIN1 are:
MAX1-MAX2 (= a) in the subtractor 701
, MIN2-MIN1 (= b) in the subtractor 702, MAX1-MIN2 (= c) in the subtractor 703, and MAX2-MIN in the subtractor 704.
1 (= d) is calculated. Next, the calculated a, b, c, and d are divided by the dividers 705 and 706 into a / d,
b / c is obtained, and the comparator 707 is obtained similarly to the circuit of FIG.
The magnitude relationship is compared at, and the larger value is selected by the switch 708. The selected value is output to comparator 709
Is compared with a certain threshold value TH to determine whether or not separation is performed. That is, in the circuit of FIG. 2, the judgment is made based on the contrast difference, whereas in this judgment circuit, the judgment is made based on the contrast ratio.

Next, the determination in this example will be described based on actual image information shown in FIG. FIG. 8 shows a line image of a level "20" synthesized on a natural image. In the block shown in FIG. 8, MAX1 = 39, MAX
2 = 38, MIN2 = 31, and MIN1 = 20, which cannot be separated by the determination circuit of FIG. However, in this determination circuit, a = 1 and b = 1
1, c = 8 and d = 18 are obtained, and b / c> a / d is satisfied. Here, b / c = 1.375, for example, the threshold value TH
In the case where = 1.3, MIN1 is separated.

In this example, a divider is used for the sake of simplicity. However, in actuality, the comparison of b / c> a / d is replaced with the comparison of b × d> a × c using the divider. It may be replaced, or even if a divider is used, the numerator may be bit-shifted so that no decimal point appears.

This will be described in more detail based on the histogram shown in FIG. When the histogram of one block is divided into two as shown in FIG. 9, this example is effective when the contrast of the level of the natural image in the block is small. That is, even if b (MIN2-MIN1) in FIG. 9 is the same, the contrast (c (MAX1-MI
N2) is large and small, MI
The importance (conspicuousness) of N1 differs. That is, it can be used to determine whether or not to separate the ratio of the contrast (c) in the natural image and the contrast (b) of the level considered to be a character or line drawing portion from the level closest to the level of the line image in the natural image. it can.

[0038]

Next, another embodiment of the present invention will be described in detail with reference to the drawings.

FIG. 10 shows another embodiment applied to a color image compression apparatus. Now, for example, color image information is represented by R (red), G (green), and B (blue).
It is assumed that image information is input on the three planes. Since the input R, G, and B have a strong correlation between the colors, color space conversion is performed on, for example, Y (luminance), U, and V (both color differences) to reduce the correlation.

In FIG. 10, reference numeral 1001 denotes a means for automatically determining the separation of the Y component, as described in the above-described embodiment.
It is determined whether or not to perform automatic separation using the four values of MAX2, MIN2, and MIN1. In this embodiment, for example, when a bitmap 1002 in which a pixel considered to be a character part of a Y component is set to “1” by satisfying a determination condition is created, other bitplanes of the same block are separated using this bitmap. It is to confirm whether it is OK to do so. This check determines whether the pixels of the bitmap “1” take the same value (called the same value) in both the U component and the V component planes (the U component same value checking means 1003 and the V component same value check). Means 1004). And 100
Reference numeral 5 denotes an AND circuit, which determines whether the “1” pixel in the bitmap 1002 has the same value in both the U component plane and the V component plane, and transmits the same to the separation determining unit 1006.

Here, a description will be given using actual image information. (A) to (c) of FIG. 11 are diagrams each showing information of one block in the same spatial coordinates of a certain image, and correspond to a Y plane, a U plane, and a V plane.
When the separation is determined by the separation circuit 106 in FIG.
The pixel portion of MIN1 “20” is separated, and the created bitmap is as shown in FIG. 11D. Using FIG. 11D, the equivalence check of the U plane and the V plane is performed. In (b) of FIG. 11, all of the “1” pixel portions in (d) of FIG. 11 satisfy the equivalence condition with a value of “130”. It does not matter whether or not the value of “130” appears in other pixels. If there is a pixel having a value other than “130” even though the bitmap is “1”, The equivalence condition will not be satisfied. Also, in (c) shown in FIG.
(D) of the pixel shown in (d) satisfies the same value condition with the value of "108". That is, a pixel of a certain color value artificially generated by a computer or the like is:
Even in each of the Y, U, and V planes, each plane can have the same pixel value. If this is confirmed on the other two planes based on the bitmap that has been successfully separated on a certain plane, it can be determined that the value is not one color if the equivalence condition is not satisfied.

In FIG. 10, in addition to the automatic separation judgment of the Y component, the automatic separation judgment of the U component and the automatic separation judgment of the V component are also performed. The same value checking means is applied to other color components other than the determined plane. Then, the separation determining means 1006
Inputs a signal indicating whether or not to separate each color component, and determines the separation. This determination is made by giving a priority such as separation in the Y component → separation in the U component → separation in the V component.

As described above, in order to determine whether or not blocks in the same space can be separated by being divided into color components, for example, even in the case where separation is performed based on a contrast difference, the vicinity of a certain point in a color space can be determined. It is possible to hold down. In other words, when separation is not possible with this method, only when the color of the character or line image synthesized with the color of the natural image of the background is very close, that is, when there is not much deterioration in image quality even if compression / expansion is performed without separation. Becomes Although various conditions have been described above, it is needless to say that these conditions may be used alone or in combination.

The present invention may be applied to a system composed of a plurality of devices or an apparatus composed of one device. Needless to say, the present invention can be applied to a case where the present invention is achieved by supplying a program to a system or an apparatus.

[0045]

As described above, according to the present invention,
Line drawing part artificially synthesized on the natural image in one block
Is the first and second largest in the block as described above.
Value and the first and second smallest values and their relationship
It was separated consideration, on the basis of these, to identify the line drawing unit
As a result, it is possible to easily and accurately extract the line drawing part.
It becomes possible .

[Brief description of the drawings]

FIG. 1 is a schematic block diagram illustrating a configuration of an image compression device according to an embodiment.

FIG. 2 is a diagram illustrating a detailed configuration of a determination circuit according to the embodiment.

FIG. 3 is a diagram showing a comparison of DCT components with and without separation.

FIG. 4 is a diagram illustrating a case where separation is not performed.

FIG. 5 is a diagram illustrating a determination in the present embodiment using a histogram.

FIG. 6 is a diagram illustrating a detailed configuration of a determination circuit according to a first modification.

FIG. 7 is a diagram illustrating a detailed configuration of a determination circuit according to a second modification.

FIG. 8 is a diagram showing blocks that can be extracted in a second modification.

FIG. 9 is a diagram illustrating a determination in Modification Example 2 using a histogram.

FIG. 10 is a schematic block diagram illustrating a configuration of an image compression device according to another embodiment.

FIG. 11 is a diagram showing a block and a bit map of each color component in another embodiment.

──────────────────────────────────────────────────続 き Continued on the front page (58) Field surveyed (Int.Cl. ⁷ , DB name) H04N 1/41-1/419 H04N 7/30

Claims (2)

(57) [Claims] 1. An image compression apparatus for compressing image and natural image and a line image are mixed, the blocking means for blocking an input image data into predetermined blocks, obtained in the blocking means based on each block <br/> click, the image data having a predetermined pixel value, the line
Extracting means for extracting an image as an image ;
And a compression means for compressing the image data after replacing the pixel values, the extraction means is a maximum value (MAX1 and in the block
Yes ), minimum value (assumed to be MIN1), second largest value
(MAX2) The second smallest value (MIN2)
4) are extracted, and MAX1-MAX2, MI
N2-MIN1 is calculated, and the subtracted values are compared with each other and a predetermined threshold is calculated.
MAX1 or MIN1 depending on the comparison result with the value.
An image compression apparatus for determining which one to extract . 2. An image in which a natural image and a line image are mixed is compressed.
An image compression apparatus for compressing input image data into predetermined blocks.
On the basis of each block obtained by the blocking means and the blocking means,
Extract image data with a predetermined pixel value as a line image
Extracting means, and a pixel value of the image data extracted by the extracting means,
Compression means for compressing the image data after replacement with pixel values;
And the extraction means has a maximum value (MAX1), a minimum value (MIN1), a second largest value (MAX2), and a second smallest value (MIN2) in the block. de disconnect the four values of, (MAX1-MAX2) /
(MAX2-MIN1) , (MIN2-MIN1) /
(MAX1-MIN2) is calculated, the divide value and between
MAX1 or M according to the result of comparison with the predetermined threshold
An image compression apparatus for determining which one of IN1 is to be extracted.