JPH08331391A - Image processing unit and method therefor - Google Patents

Abstract

PURPOSE: To provide the image processing unit and its method in which an area of an input image is correctly divided based on a characteristic of the image and coding in response to a characteristic is applied to the divided areas. CONSTITUTION: A background decision section 5 decides a state of a background of an input image. An adaptive quantization section 6 binarizes an input image based on a state of a background, binarizes the input image subject to edge emphasis and integrates two binarized results to generate a background binary image. An area divider 7 applies area division to the input image based on the background binary image. The divided area is coded by a multi-value image coder 9 or a binary image coder 10 in response to the characteristic.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image processing apparatus and a method thereof, and more particularly, to an image processing apparatus and a method thereof for dividing an image into regions and encoding them.

[0002]

2. Description of the Related Art In recent years, with the rapid progress of image coding technology, international standardization has progressed, and the JPEG system has been used as a lossy coding for multivalued color images, and the JPEG system has been used as a lossless coding for black and white binary images. The JBIG method has been standardized. Furthermore, the standardization of new coding methods is also in progress.

FIG. 1 is a block diagram showing an example of the configuration of an image coding apparatus. 1001 is a scanner for inputting an image, 1002 is a selector for selecting an output destination of the input image according to a user's instruction, and 1003 is a JPEG code for the input image. JPEG encoder to convert, 10
04 is a binarizer for binarizing the input image, 1005 is a JBIG encoder for encoding the binary image, 1006 is a selector for selecting the input destination of the code according to the user's instruction, 1007 is a communication interface, 1008 Is a communication line such as a telephone line or a LAN.
In addition, the selector 1002 and the selector 1006 operate in synchronization,
The image coded through the selector 1002 is the selector 10
It is sent to the communication interface 1007 via 06.

In such a configuration, when the user wants to transmit a color image, first, the JPEG encoder 1003 is used as an output destination of the selector 1002 and an input destination of the selector 1006.
Select Next, the scanner 1001 is operated to input the image data, and the JPEG encoder 1003 is input via the selector 1002.
Image data is input to and the coded data encoded by the JPEG method is obtained. This code data is input to the communication interface 1007 via the selector 1006, and then sent out to the communication line 1008 by the communication interface 1007 according to a predetermined protocol.

When transmitting a black and white binary image, the user first selects the binarizer 1004 as the output destination of the selector 1002 and the JBIG encoder 1005 as the input destination of the selector 1006. To do. Then the scanner 10
01 is operated to input image data and input to the binarizer 1004 via the selector 1002. The binarizer 1004 binarizes the image data by comparing a preset threshold value with the input image data. The binarized image data is JB
The code data is input to the IG encoder 1005 and encoded by the JBIG method. This code data is input to the communication interface 1007 via the selector 1006, and then the communication interface 1007 uses the communication line 1008 according to a predetermined protocol.
Sent to

[0006]

However, the above-mentioned technique has the following problems. In other words, in the image coding apparatus that switches the coding method for each image, the characters and line drawings included in the image are also multi-value coded by the JPEG method, etc. As a result, there is a drawback that the quality of characters is significantly deteriorated.
Further, when the binary encoding is performed by the JBIG method or the like, there is a drawback that the gradation part included in the image becomes flat by the binarization.

As an improved JPEG system, a method of judging the image feature for each coding unit (8 × 8 pixels) and switching the quantization coefficient or switching the coding table has been proposed. Even in this method, the number of reference pixels for determining the characteristics of the image is small, and a correct determination result may not be obtained.

A method of binarizing a multi-valued image and separating a character / line drawing area from an area such as a photograph may be considered to globally determine the image area and switch the encoding method. The method of extracting an image area having a predetermined density or higher using a fixed threshold value or a variable threshold value has a drawback in that it cannot deal with low-contrast characters / line drawings and blank characters.

Further, when binarizing a multi-valued image, if the edges of characters, line drawings, etc. are originally blunt, the obtained binary image becomes thick, which may be a cause of erroneous determination when separating areas. Become. In particular, when there is no space between characters and the character area is erroneously determined to be the gradation area, the character / line drawing area is multi-value encoded, and the encoding method is switched for each image. It becomes the same as the image encoding device.

The present invention is intended to solve the above-mentioned problem. For example, an input image is correctly divided into regions based on the characteristics of the image, and the divided regions are encoded according to the characteristics. An object of the present invention is to provide an image processing apparatus and a method therefor.

[0011]

[Means for Solving the Problems] and

The present invention has the following structure as one means for achieving the above object.

An image processing apparatus according to the present invention includes a determination means for determining the state of a background of an input image, a correction means for correcting the input image, and a quantum processing for the input image based on the determination result of the determination means. And quantizing the image corrected by the correcting means, and integrating the two quantization results, a dividing means for dividing the input image into regions based on the integrated quantization result, and dividing Coding means for coding each of the generated regions according to the characteristics of the region.

Further, a judging means for judging the background condition of the input image, a correcting means for correcting the input image, and a first quantizing means for quantizing the input image based on the judgment result of the judging means. A second quantizing means for quantizing the image corrected by the correcting means on the basis of the judgment result, and based on the judgment result, the input image and the image corrected by the correcting means, Third quantizing means for quantizing the input image, and integrating means for integrating the respective quantization results obtained by the first to third quantizing means,
It is characterized in that it has a dividing means for dividing the input image into areas based on the integrated quantization result, and an encoding means for encoding each divided area according to the characteristics of the area.

According to the image processing method of the present invention, a determination step of determining a background state of the input image, a correction step of correcting the input image, and a quantization of the input image based on a determination result of the determination step. A first quantization step, a second quantization step of quantizing the image corrected in the correction step based on the determination result, and each quantization obtained in the first and second quantization step An integration step of integrating the results, a division step of dividing the input image into areas based on the integrated quantization result, and an encoding step of encoding each of the divided areas according to the characteristics of the area. Is characterized by.

Further, a determination step of determining a background state of the input image, a correction step of correcting the input image, and a first quantization step of quantizing the input image based on the determination result of the determination step. A second quantization step of quantizing the image corrected in the correction step based on the determination result, and the input image based on the determination result, the input image and the image corrected in the correction step. A third quantizing step, a integrating step of integrating the respective quantization results obtained in the first to third quantizing steps, and a region of the input image based on the integrated quantizing result. It is characterized by including a dividing step of dividing and an encoding step of encoding each divided area according to the characteristics of the area.

[0016]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An image processing apparatus according to an embodiment of the present invention will be described in detail below with reference to the drawings.

[0017]

First Embodiment [Structure] FIG. 2 is a block diagram showing an example in which an image processing apparatus according to an embodiment of the present invention is applied to an image transmitting apparatus.

In FIG. 1, reference numeral 3 denotes a CPU, which controls the entire apparatus via a bus 4 in accordance with a program stored in a built-in ROM, the storage device 2, or the like. Image data and control signals are exchanged via the bus 4.

Further, 1 is a scanner for inputting an image, 2 is a storage device for storing image data and the like, 5 is a background determining section for determining the background state of the input image and outputting background information, 6
Is an adaptive quantizer that appropriately quantizes the input image, 7 is a region divider that separates regions with similar image characteristics based on the quantization result, and 14 is the region division result output from the region divider 7. A region division memory for storing a region information encoder, 8 is a region information encoder for encoding the region division result, 9 is a multivalued image encoder for encoding a multivalued image such as a photograph included in the input image, 10
Is a binary image encoder that binarizes and encodes a binary image such as a character or line drawing included in the input image, 11 is a synthesizer that synthesizes and outputs code data output from each encoder, 12
Is a communication interface (communication I / F), 13 is a telephone line or LAN
Is a communication line such as.

In the following, the input image is a black-and-white multivalued image of 8 bits / pixel, and the encoding method of the multivalued image encoder 9 is JPEG.
The coding method is JB.
Although the IG encoding method will be described, the present invention is not limited to these.

[Background Determining Unit] FIG. 3 is a block diagram showing a configuration example of the background determining unit 5.

In the figure, reference numeral 21 denotes a frame memory, which stores input image data for one screen. Reference numeral 22 denotes a flat pixel extractor, which sequentially reads the target pixel and the peripheral pixels from the frame memory 21, and outputs the pixel value when the target pixel is a pixel forming a flatness. As the flat pixel determination of the flat pixel extractor 22, a known method, for example, a minimum / maximum value extraction method or a “binarization processing method for a document in which a binary image and a grayscale image are mixed” (Tetsuya, Ochi, IEICE) Journal 1984/7 Vo
l.J67-B No.7 pp.781-788).

Reference numeral 23 is a frequency counter composed of 256 counters. Each counter corresponds to each input value, and the counter corresponding to the pixel value output from the flat pixel extractor 22 counts up. A maximum frequency extractor 24 outputs a value (0 to 255) corresponding to the counter having the maximum count value (frequency) in the frequency counter 23. Reference numeral 25 is a maximum frequency density range extractor, which displays each count value of the frequency counter 23 in hist [i]
(i = 0 to 255), the value output from the maximum frequency extractor 24 is b
(Hereinafter, referred to as “background candidate value”), as shown in FIG. 4, as an example, in the vicinity of the background candidate value b, the frequency is smaller than a preset threshold Th1 and the background candidate value b is the highest. Pixel values bt1 (base candidate upper limit) and bt0 (base candidate lower limit) that are close to each other are obtained. The background candidate lower limit value bt0 and the background candidate upper limit value bt1 are the threshold values Th near the background candidate value b.
It can be obtained from the intersection of the straight line of 1 and the frequency curve.

Further, 37 is a differencer, which is the lower limit value bt0 of the background candidate.
And the difference w (background density range) between the background candidate upper limit value bt1.
A background occupancy calculator 26 is a frequency counter 2 corresponding to the pixel value sandwiched between the background candidate lower limit value bt0 and the background candidate upper limit value bt1.
The value obtained by adding each count value of 3 is divided by the number of pixels of the entire image to obtain the occupancy ratio of the background. 27 and 28 are comparators,
The comparator 27 compares the threshold Th2 with the background density width w, and the comparator 28
Compares the threshold value Th3 with the background occupancy rate and outputs '1' if the input value is larger than the threshold value, and outputs '0' otherwise. 29 is an inverter that inverts the output of the comparator 27, and 30 is an AND that ANDs the output of the inverter 29 and the output of the comparator 28.
It is a gate.

Next, the operation of the background determination section 5 will be described.

Prior to the start of processing, the CPU 3 clears the frame memory 21 and the frequency counter 23. Then, CPU3
The scanner 1 or the storage device 2 according to the user's instruction.
The black-and-white multivalued image read from is stored in the frame memory 21. When the image for one screen is stored in the frame memory 21, the flat pixel extractor 22 sequentially reads out the pixel of interest and eight pixels (reference pixels) adjacent thereto from the frame memory 21,
Find the maximum difference between the value of the pixel of interest and the value of each reference pixel,
When the difference value is smaller than the threshold value Th4, the pixel of interest is output as a pixel that constitutes flatness. The frequency counter 23 counts up the counter corresponding to the output pixel value.

When the flat pixel extraction processing is completed for all the pixels of the image stored in the frame memory 21, the maximum frequency extractor 24 corresponds to the counter having the maximum count value in the frequency counter 23. Value, that is, the background candidate value b is output. Subsequently, the maximum frequency density width extractor 25 obtains the background candidate lower limit value bt0 and the background candidate upper limit value bt1 from the count content of the frequency counter 23. From the obtained background candidate lower limit value bt0 and background candidate upper limit value bt1, the subtractor 37 obtains the background density width w, and the background density width w is calculated by the comparator 27 as the threshold Th2.
Compared to.

On the other hand, the background occupancy calculator 26 obtains the background occupancy from the background candidate lower limit value bt0 and the background candidate upper limit value bt1,
The background occupancy is compared with the threshold Th3 by the comparator 28.

The output of the comparator 27 is inverted by the inverter 29 and input to the AND gate 30 together with the output of the comparator 28, and the logical product bg (background determination flag) is output.

That is, when the background density width w is smaller than the threshold value Th2 and the background occupancy rate is larger than the threshold value Th3, the background determination flag bg becomes "1" to represent "a plain background".
If this condition is not satisfied, the background judgment flag bg
Becomes "0" and represents "a background with a pattern".

[Adaptive Quantization Unit] FIG. 5 shows the adaptive quantization unit 6
3 is a block diagram showing a configuration example of FIG.

Reference numeral 41 is a frame memory for storing an input image, 42
Is an edge enhancement circuit for correcting the edge of the image, and performs edge enhancement based on the pixel of interest read from the frame memory 41 and its peripheral pixels. If the target pixel D is a pixel D ′ that has been edge-emphasized, the edge enhancement is expressed by the following equation, for example. D '= D + (4D-A1-A2-A3-A4) (1) where D = X (i, j) A1 = X (i-1, j-1) A2 = X (i + 1, j-1) A3 = X (i-1, j + 1) A4 = X (i + 1, j + 1) i, j: Coordinate value

Reference numerals 45 to 48 are comparators, and the comparator 45 is
The target pixel D input from the frame memory 41 is compared with the background candidate upper limit value bt1 input from the background determination unit 5,
If bt1> D, output '1', otherwise output '0'. The comparator 46 compares the pixel of interest D with the background candidate lower limit value bt0, and if D> bt0, the value is “1”, otherwise the value is “0”.
Is output. The comparator 47 receives the edge enhancement result D ′ of the pixel of interest input from the edge enhancement circuit 42 and the background candidate upper limit value bt.
Compare with 1 and output '1' if bt1> D ', otherwise output' 0 '. The comparator 48 compares the edge enhancement result D ′ with the background candidate lower limit value bt0, and if D ′> bt0, then outputs “1”.
Otherwise, '0' is output.

Further, 49 and 50 are AND gates respectively, and 51 is O.
An R gate 52 is a frame memory that stores the output of the OR gate 51.

Next, the operation of the adaptive quantizer 6 will be described.

When the processing of the background determination section 5 is completed, the CPU 3 accumulates the image data sequentially read from the frame memory 21 of the background determination section 5 in the frame memory 41, and when the image data for one screen is accumulated, the target pixel Are sequentially read and the comparators 45 and 46 are made to perform comparison. The comparison result of both comparators is AN
ANDed by D-gate 49. The output of the AND gate 49 is '1' if the target pixel D is between the background candidate upper limit value bt1 and the background candidate lower limit value bt0, and is '0' otherwise.

On the other hand, the edge emphasizing circuit 42 performs the operation shown in the equation (1) on the input target pixel D and the four surrounding pixels A1 to A4, and outputs the edge emphasizing result D '. Edge enhancement results
D ′ is input to the comparators 47 and 48 and compared with the background candidate upper limit value bt1 and the background candidate lower limit value bt0, respectively. The AND gate 50 ANDs the comparison results of both comparators. AND gate
In the output of 50, the pixel D'after edge enhancement is the background candidate upper limit value bt
It is '1' if it is between 1 and the background candidate lower limit value bt0, and '0' otherwise.

The outputs of the AND gates 49 and 50 are logically ANDed by the OR gate 51. That is, the OR gate 51 outputs the data “1” for the pixel sandwiched between the background candidate lower limit value and the upper limit value, and the result is stored in the frame memory 52. Therefore, the background binary image represented by the data “1” is stored in the frame memory 52.

[Region Divider] When the processing of all pixels stored in the frame memory 41 by the adaptive quantizing unit 6 is completed, the CPU 3 determines that the background determination flag bg is "1" (plain background). , Activate the area divider 7. Area divider 7
Reads the data stored in the frame memory 52,
Based on the characteristics of the image, a rectangular area such as “text area”, “photograph area”, “line drawing area”, “separator” is extracted, and the position coordinates (for example, the upper left xy coordinate) of that area, the size of the rectangular area, and the area The area information including the code indicating the determination result is output. This area information is stored in the area division memory 14.

Further, the CPU 3 sets the background judgment flag bg to "0".
In case of (base with pattern), the area divider 7 is not operated and the position coordinates of the rectangular area are (x, y) = (0,0) (origin)
The area division memory 14 stores area information including the size of the entire image as the size of the rectangular area and the code of the “photo area” as the code indicating the determination result of the area. That is,
Region division is not performed for an image whose background has a pattern.

As an image area dividing method, there are two threshold values for an image, which are disclosed in Japanese Patent Application Laid-Open No. 62-226770 “Image Area Separating Device” and Japanese Patent Application Laid-Open No. 3-126181 “Document image area dividing method”. A method of binarizing and determining a character area, a photograph, a table area, etc. from a block of pixels, and separating the character and pattern areas as a white background area disclosed in JP-A-4-248766 "Image Area Separation Method" There are ways to do it.

[Sending of Image Data] After storing the area information in the area division memory 14, the CPU 3 uses the background candidate value b output from the background determination unit 5 via the synthesizer 11 and the communication I / F 12. , To the communication line 13.

Subsequently, the CPU 3 sequentially reads the area information from the area division memory 14, inputs the read area information to the area information encoder 8, and sends the encoded area information to the synthesizer 11. Next, if the rectangular area represented by the area information is a "photograph area", the image data of the area read from the frame memory 21 of the background determination unit 5 is input to the multi-valued image encoder 9, and JPEG The encoded image data is sent to the synthesizer 11. If the read area information is any one of "character area", "line drawing area", and "separator", the adaptive quantization unit
The image data (represented by data "0") of the area read from the frame memory 52 of 6 is the binary image encoder 1
Input to 0 and input the JBIG encoded image data to the synthesizer 11
Send to.

The synthesizer 11 synthesizes the coded area information and the coded data obtained by coding the image data of the rectangular area represented by the area information, and the communication interface 12
To the communication line 13 via.

[Summary] As described above, in this embodiment, the base is determined based on the frequency distribution of the pixel values of the input image to be encoded, the base is extracted, and the "plain base" or "pattern is provided". It is determined whether or not it is a “base”. Then, when the background of the input image is plain, a binary image is formed from the input image and the edge-enhanced input image based on the density of the extracted background, and a character / line drawing area and an area such as a photograph are formed. And the image data of the area is encoded by an encoding method suitable for the separated area. If the background of the input image has a pattern,
The entire image is encoded by the multi-valued image encoding method without performing region division. Therefore, according to this embodiment, the following effects can be obtained.

(1) Since the background is extracted from the distribution of pixel values, it is possible to suppress the influence of the show-through which allows the back side of the document to show through when the background density is low. (2) Based on the extracted background density , Since a binary image is formed from the input image and the edge-corrected input image to separate the regions, it is possible to correctly separate low-contrast characters / line drawings and outline characters, as well as the edges of characters and line drawings. Even if it is originally blunt, the binarized image does not become thick, and it is possible to prevent erroneous judgment when separating areas. (3) The image of the area is encoded by a coding method suitable for the area. Since the data is encoded, the edges, which are the features of characters and line drawings, are not blunted, and the quality of characters is not significantly reduced, and the gradation part does not become flat due to binarization. (4) The background is patterned , Then the entire image Since the image is encoded by the value image encoding method, it is possible to perform the optimal encoding including the background pattern. In addition, the input image and the edge-corrected input image are binarized, and the binarized result is obtained. The area extraction method of the present embodiment, which extracts an area using an integrated binary image, provides high area determination accuracy and also extracts a character / line drawing area appropriately for an image that is negative-positive inverted. can do.

[0047]

Second Embodiment An image processing apparatus according to the second embodiment of the present invention will be described below. It should be noted that in the second embodiment, substantially the same configurations as those in the first embodiment are designated by the same reference numerals, and detailed description thereof will be omitted.

[Structure] FIG. 6 is a block diagram showing an example in which the image processing apparatus of the second embodiment according to the present invention is applied to an image transmitting apparatus. The difference from the first embodiment shown in FIG. This means that the area image division result and the binary image are input to the value image encoder 10, and the multi-valued image is also input.

In the following, the input image is a color multi-valued image of 8 bits / pixel for each RGB color, the multi-valued image encoder 9 uses the DPCM encoding method, and the binary image encoder 10 uses the color image. The encoding method will be described as an MMR encoding method, but the present invention is not limited to these.

[Background Determining Section] FIG. 7 is a block diagram showing a configuration example of the background determining section 5.

A luminance generator 122 converts the input RGB image data into a luminance signal L * in the CIE1976 L * a * b * uniform color space. A frame memory 123 stores L * image data for one screen. A smoothing device 126 reads image data from the frame memory 123 and smoothes 3 × 3 pixels. Reference numeral 124 is a minimum value calculator, and as shown in one example in FIGS. 8A and 8B,
Pixel value bv1 (second background candidate upper limit value) that shows a local minimum value in the vicinity of the background candidate value b and is closest to the background candidate value b
And bv0 (lower limit value of second background candidate) are obtained. In the following, the output bt0 of the maximum frequency density range extractor 25 is referred to as a first background candidate lower limit value, and bt1 is referred to as a first background candidate upper limit value.

Next, the operation of the background determining section 5 will be described focusing on the points different from the first embodiment.

Prior to the start of processing, the CPU 3 clears the frame memory 21 and the frequency counter 23. Then, CPU3
The scanner 1 or the storage device 2 according to the user's instruction.
The color multi-valued image read from is stored in the frame memory 21. When the image for one screen is stored in the frame memory 21, the luminance generator 122 sequentially reads the RGB image data from the frame memory 21, and writes the converted L * image data in the frame memory 123.

L * image data for one screen is the frame memory
Once stored in 123, the smoother 126
The target pixel and its surrounding pixels (3 × 3 pixels) are read out from 3, and the target pixel is smoothed. The flat pixel extractor 22 of the present embodiment determines whether or not the target pixel is a flat pixel from the smoothed target pixel and the eight reference pixels adjacent thereto.

When the smoothing process and the flat pixel extraction process are completed for all the pixels stored in the frame memory 123, the maximum frequency extractor 24 outputs the background candidate value b, and the maximum frequency density range extractor. 25 is the lower limit value of the first base candidate bt
The minimum value calculator 124 outputs 0 and the first background candidate upper limit value bt1, and the second background candidate lower limit value bv0 and the second background candidate upper limit value bv1.

[Adaptive Quantization Unit] FIG. 9 shows the adaptive quantization unit 6
3 is a block diagram showing a configuration example of FIG.

Reference numerals 147 and 148 denote selectors, which select inputs according to the background judgment flag bg and output them. Selector 148
Selects the first or second background candidate lower limit value and inputs the selected background candidate lower limit value to the comparators 46 and 48. selector
147 selects the first or second background candidate upper limit value and inputs the selected background candidate upper limit value to the comparators 45 and 47.

Reference numerals 149 to 152 denote comparators, respectively.
Compares the background candidate value b with the edge-enhanced pixel value output from the edge emphasizing circuit 42, and if the pixel value after the edge emphasis is larger than the background candidate value b, '1', otherwise. '0' is output. The comparator 150 compares the background candidate value b with the pixel value output from the frame memory 41, and outputs '1' if the background candidate location b is larger than the pixel value, and outputs '0' otherwise. To do. The comparator 151 compares the background candidate value b with the pixel value output from the frame memory 41,
If the pixel value is larger than the background candidate value b, '1' is output, otherwise '0' is output. The comparator 152 uses the background candidate value b
And the pixel value after edge enhancement are compared, and if the background candidate value b is larger than the pixel value after edge enhancement, '1' is output, and otherwise, '0' is output.

153 and 154 are AND gates, respectively. The AND gate 153 logically ANDs the outputs of the comparators 149 and 150, and the AND gate 154 logically ANDs the outputs of the comparators 151 and 152, and outputs it to the OR gate 51.

Next, regarding the operation of the adaptive quantizer 6,
The description will focus on the parts that differ from the first embodiment.

When the processing of the background determination unit 5 is completed, the CPU 3 sequentially reads from the frame memory 123 of the background determination unit 5.
When the L * image data is stored in the frame memory 41 and the image data for one screen is stored, the pixel of interest is sequentially read out and the comparators 45 to 48 and 149 to 152 are made to perform comparison. The comparison results of these comparators are ANDed by AND gates 49, 50, 153 and 154, respectively, and the ANDed results are ORed by OR gate 51 and written to frame memory 52.

Here, when the background determination flag bg is "1" (background with a pattern), the selector 147 determines the first background candidate upper limit value b.
Since t1 is selected and the selector 148 selects the first background candidate lower limit value bt0, the operations of the comparators 45 to 48 are the same as in the first embodiment. That is, the output of the AND gate 49 is '1' if the pixel of interest D is between the first background candidate upper limit value bt1 and the first background candidate lower limit value bt0, and is '0' otherwise. AND gate 50
In the output of, the pixel D'after edge enhancement is the first background candidate upper limit
It is '1' if it is between bt1 and the first background candidate lower limit value bt0, and '0' otherwise.

When the background determination flag bg is "0" (plain background), the selector 147 selects the second background candidate upper limit value bv1 and the selector 148 selects the second background candidate lower limit value bv0. Therefore, the output of the AND gate 49 is '1' if the pixel of interest D is between the second background candidate upper limit value bv1 and the second background candidate lower limit value bv0, and is '0' otherwise. The output of the AND gate 50 shows that the pixel D'after edge enhancement is the second background candidate upper limit value bv.
If it is between 1 and the second background candidate lower limit value bv0, it is '1', and if not, it is '0'.

The outputs of the AND gates 153 and 154 are "1" if the background candidate value b is between the target pixel D and the edge-enhanced pixel D ', and "0" otherwise. .

That is, the OR gate 51 outputs the data "1" for the pixel between the lower and upper limit values of the background candidate and the pixel between the values before and after the edge enhancement between the candidate background value b. Therefore, the background binary image represented by the data “1” is stored in the frame memory 52. However, when the background has a pattern, the range of the background candidate value is wider than when the background is plain, so that the background including the pattern is extracted.

[Region Divider] When the processing of all the pixels stored in the frame memory 41 by the adaptive quantizing unit 6 is completed, the CPU 3 differs from the first embodiment in that the background judgment flag bg
Regardless of, the area divider 7 is activated. The area divider 7 is
The data stored in the frame memory 52 is read, rectangular areas such as “character area”, “photograph area”, “line drawing area”, and “separator” are extracted based on the characteristics of the image, and the position coordinates of that area (for example, the upper left corner) are extracted. Area information including the (xy coordinates), the size of the rectangular area, and the code indicating the area determination result is output. This area information is stored in the area division memory 14.

[Sending of image data] When all the pixels stored in the frame memory 41 are processed by the adaptive quantizing unit 6 and the area information is stored in the area dividing memory 14, the CPU 3 determines the background. Substrate candidate value b output from section 5
To the communication line 13 via the synthesizer 11 and the communication I / F 12.

Subsequently, the CPU 3 sequentially reads the area information from the area division memory 14, inputs the read area information to the area information encoder 8, and sends the encoded area information to the synthesizer 11. Next, if the rectangular area represented by the area information is a "photograph area", the image data of the area read from the frame memory 21 of the background determination unit 5 is input to the multi-valued image encoder 9, and the DPCM is used. The encoded image data is sent to the synthesizer 11. If the read area information is any one of "character area", "line drawing area", and "separator", the adaptive quantization unit
The image data of the area (represented by data “0”) read from the frame memory 52 of 6 and the image data of the area read from the frame memory 21 of the background determination unit 5 to the binary image encoder 10. Input, MMR encoded image data and foreground color information (details will be described later) 11
Send to.

The synthesizer 11 synthesizes the coded area information, the coded data obtained by coding the image data of the rectangular area represented by the area information, the foreground color information, and the like, and via the communication interface 12. It is sent to the communication line 13.

[Binary Image Encoder] FIG. 10 is a block diagram showing a configuration example of the binary image encoder 10.

In the figure, reference numeral 161 denotes a latch, which latches RGB image data sequentially input from the frame memory 21 of the background determination section 5 and synchronizes with the RGB image data from the frame memory 52 of the adaptive quantization section 6. When the input binary image data is '0' (other than the background), the latched data is output.

A frequency memory 162 is composed of a memory for each of the RGB component colors, and a latch 16 is provided at the address terminals thereof.
1 to R, G, B data are input respectively. An adder 163 adds 1 to the input data and outputs the result. That is, when RGB image data is input to the frequency memory 162, its R, G, B
The data stored in the memory is output using the values as addresses. The data output from the frequency memory 162 is added with 1 by the adder 163, and then, again,
It is stored in the frequency memory 162. That is, the frequency memory 162 stores the respective distributions of R, G, B data of image data excluding the background. The frequency memory 162 is
Cleared by CPU3 prior to encoding.

Reference numeral 164 denotes a maximum frequency extractor, which outputs each address stored in the frequency memory 162, at which each RGB frequency is maximum, as RGB data. Below, this frequency is maximum
RGB data is called the "foreground color" of the input image. Reference numeral 165 is an MMR encoder, which performs MMR encoding of the binary image data input from the frame memory 52 of the adaptive quantizer 6.

Now, the read area information is "character area".
If it is either "line drawing area" or "separator", CPU3
Is the binary image data (represented by data “0”) of the area from the frame memory 52 of the adaptive quantizing unit 6 and the RGB multi-value image data of the area from the frame memory 21 of the background determination unit 5. The data is synchronously read and input to the binary image encoder 10.

When the frequency count process is completed for all the pixels in the area, the maximum frequency extractor 164
The foreground color is extracted from the count result stored in the frequency memory 162 and output to the synthesizer 11. Subsequently, the CPU 3 reads the binary image data in the area from the frame memory 52,
The MMR encoder 165 performs MMR encoding and outputs the encoded data to the synthesizer 11.

[Summary] As described above, according to this embodiment, the same effect as that of the first embodiment can be expected.
Since two sets of background density lower limit value and background density upper limit value are set, even if the background of the input image has a pattern, it is possible to accurately determine the area, and a method suitable for the characteristics of each divided area Can be encoded.

Further, when the pixel value is corrected by sandwiching the background candidate value (background density) by the edge enhancement processing, for example, the input pixel value is larger than the background candidate upper limit value, and the pixel value after the edge enhancement is the background candidate. Even if it is smaller than the lower limit value, the input image can be appropriately binarized.

Although an example in which a luminance signal is generated from a color image and area division is performed has been described above, the present invention is not limited to this, and a G signal may be used in place of the luminance signal. The same operation as above may be performed. Further, CIE1976 L * u * v * or YIQ may be used as the uniform color space. Further, the area may be determined by prescanning, and then the color image may be read.

[0079]

Third Embodiment An image processing apparatus according to the third embodiment of the present invention will be described below. In addition, in the third embodiment, substantially the same configurations as those in the first embodiment are designated by the same reference numerals, and detailed description thereof will be omitted.

[Structure] FIG. 11 is a block diagram showing an example in which the image processing apparatus of the third embodiment according to the present invention is applied to an image transmitting apparatus.

The present embodiment is an example in which the encoding process described in the first and second embodiments is performed by software stored in the program memory 202 based on a user's instruction input from the terminal 203. is there. The CPU 3 includes a sufficient work memory necessary for this encoding process.

In this embodiment, the input image is 8 bits /
A black-and-white multi-valued image of pixels, and the JPEG encoding method
The encoding method will be described as a binary image encoding method as a JBIG encoding method, but the present invention is not limited to this.

[Processing Procedure] FIGS. 12 to 15 are flowcharts showing an example of the encoding process, which is executed by the CPU 3 in accordance with the program stored in the program memory 202 when the terminal 203 instructs the encoding. Is.

Initialization When the image data is read from the storage device 204 in step S1 and the reading of the image data for one screen is completed, step S2
Then, set the number of areas count = 1, and as the first record of the area information table, set (0,0) as the "position coordinate" that represents the upper left corner of the rectangular area, and the A code corresponding to unseparated is stored as the image size, which is a “code” indicating the determination result of the area. Fig. 16
A shows an example of the contents of the area information table when the process of step S2 is completed, and the “area size” is 3,072 ×
An example in which 4,096 pixels are stored is shown.

Subsequently, the value of count is determined in step S3, and if count ≠ 0, the process proceeds to step S4, and all histogram tables hist [i] (i = 0 to 255) are cleared.

Extraction of Flat Pixels Next, the extraction process of flat pixels is performed.
In S5, it is determined whether or not the flat pixel extraction processing has been completed for all the pixels in the count area of the rectangular area recorded in the area table. And if there are unprocessed pixels, step
In step S6, the image data of the rectangular area recorded in the count-th area of the area information table is sequentially read, and the value sm indicating the maximum absolute value of the difference between the target pixel p and its surrounding pixels xi is obtained. Next, in step S7, the maximum difference value sm is compared with the threshold value Th7, and if sm ≧ Th7, the process proceeds to step S9, sm <Th7
If so, the pixel of interest p is regarded as a pixel forming a flatness, and hist [p] of the histogram table is incremented in step S8. Then, after moving the target pixel to the next pixel in step S9, the process returns to step S5.

Base extraction Steps S6 to S for all pixels in the countth area
When the process of 9 is completed, the process proceeds to step S10, and hist [i] (i = 0 to 2 indicating the maximum value (frequency) in the histogram table is displayed.
55) is obtained and the parameter i is set as the background candidate value b (that is, hist [b] is the maximum value). Next, in step S11, the background candidate lower limit value bt0 and the background candidate upper limit value bt1 are obtained, and further, the background density width w is obtained by the following equation. w = bt1-bt0… (2)

Then, in step S12, the background candidate lower limit value b
All hist [i] included in the range from t0 to the base candidate upper limit value bt1 are added and divided by the number of pixels m in the area to obtain the base occupation rate s. s = sum / m = Σhist [i] / m (3) However, Σ operation is i = bt0 ~ bt1

Subsequently, in step S13, the background occupancy s and the threshold Th2 are compared, and the background density width w and the threshold Th3 are compared, and s> Th2 and w <
If it is Th3, it is determined that there is a base, and the process proceeds to step S13,
Substitute bt0 for the base lower limit value b0 and bt1 for the base upper limit value b1.
If the background occupancy s and the background density width w do not satisfy the above conditions, it is determined that there is no background, and the process proceeds to step S014, where the minimum value of hist [i] is set to the second background candidate lower limit value bv0. The background candidate upper limit value bv1 is obtained, and in step S15 the background lower limit value b0
Substitute bv0 for and base upper limit value b1 for bv1.

Quantization Subsequently, adaptive quantization processing is performed. First, step S1
In step 7, it is determined whether or not the quantization process has been completed for all the pixels in the count area of the rectangular area recorded in the area table. Then, if there is an unprocessed pixel, the process proceeds to step S18, the image data of the rectangular area recorded in the count-th area of the area information table is sequentially read, and the pixel of interest p and the background lower limit value b
0 is compared with the background upper limit value b1. If b0 <p <b1, the binarization result n is set to '1' in step S20, and otherwise n is set to '0' in step S19.

Subsequently, in step S21, edge enhancement is performed from the target pixel and its surrounding pixels based on the equation (1) to obtain the target pixel p'after edge enhancement. And step S22
Then, the pixel of interest p ′ after edge enhancement, the background lower limit value b0 and the background upper limit value b1 are compared, and if b0 <p ′ <b1, step S24
Sets the binarization result e to '1', otherwise step S2
Set e to '0' with 3.

Subsequently, in step S25, the logical sum (n | e) of the binarization result n and the binarization result e is set to the binarization result q, and then the target pixel is moved to the next pixel in step S26. After, step S17
Return to. The obtained binarization result q is stored in the memory as a binary image.

Area division Steps S18 to S for all pixels in the countth area
When the process of S26 is completed, the process proceeds to step S27, where the binarized image obtained by the binarization result q is a rectangle such as "character region", "photograph region", "line drawing region", "separator", etc. based on the characteristics of the image. Extract a region. And step S28
Then, the variable j = 1 is set, and the newly extracted count_org pieces of area information are written into the area information table as the records of the i + 1th and subsequent records. FIG. 16B shows an example of the contents of the area information table when the process of step S28 ends.

Encoding Next, encoding processing is performed. First, it is determined whether the count_org areas extracted in step S29 have been encoded. Then, if there is an unprocessed area, the process proceeds to step S30, and the code indicating the determination result of the count + jth area in the area information table is "character""separator""linedrawing".
If it is a code representing a binary image such as "table", step S3
Go to 1.

If the area to be encoded is a binary image, foreground density extraction processing is performed in step S31. Specifically, clear all of the histogram table histfg [i], read the binary image in that area, and use the histogram table of the pixel value p of the multi-valued image whose pixel value corresponds to '0'. Increment the data (that is, histhg [p]) and set the density at which the histogram table data shows the maximum value to the foreground density.
Let f.

Subsequently, in step S32, the count + jth area information and the foreground density f are encoded and sent to the communication line 13 via the communication I / F 12. Next, in step S33, the binary image in the area is encoded by the JBIG method and sent to the communication line 13 via the communication I / F 12, and then the process proceeds to step S34.

On the other hand, if the area to be encoded is a multi-valued image, it is determined whether the number of areas count_org separated in step S35 is 1 or not.
After changing the “code” in the nt + jth area to a code indicating unseparated (see record 4 in FIG. 16C), the process proceeds to step S34.

If count_org = 1 (that is, there is no separated area), the process proceeds to step S36, the count + 1th area information is encoded, and the communication line 13 is transmitted via the communication I / F 12.
Send to. Next, in step S37, the multi-valued image of the area is encoded by the JPEG method, and the communication line 13 is transmitted via the communication I / F 12.
After sending to, go to step S34.

Then, after incrementing the variable j in step S34, the process returns to step S29, and when j> count_org, the first record from the area information table,
Delete the record in which the codes representing "character", "separator", "line drawing", and "table" are recorded, and reconstruct the region information table by filling in the deleted amount (see Fig. 16D) and count the number of regions.
Is counted (count = 1 in the case of FIG. 16D).

That is, after deleting the first record (area) in which the area is separated and the record (area) having a code other than "unseparated", the process returns to step S3 and count = 0 (area to be separated). If there is no), all processing ends. Note that, in the example of FIG. 16D, since the “unseparated” area remains, the processing from step S4 is executed again.

[Summary] As described above, according to this embodiment, the same effect as that of the above-described embodiments can be expected, and by dividing the repetitive area, the binary code is applied to the flat area in the small area. Since the encoding efficiency is improved, the character and the line drawing can be reproduced in binary at the same time, so that the character and the line drawing are not deteriorated and the image quality of the transmission image can be improved.

[0102]

[Modification] In each of the above-described embodiments, an example in which an input image is encoded and applied to an image transmitting apparatus for transmitting to a communication line has been described, but the present invention is not limited to this and the encoding is performed. It can also be applied to an image storage device that stores the input image in a storage medium such as a hard disk, an optical disk, a magneto-optical disk, or a CD-ROM.

The color space and the number of bits of the image to be coded by the present invention are not limited to those in the above-mentioned embodiment, and for example, in the color space such as CIE1976 L * a * b *, YMCK, HSL, etc. , Any bit depth such as 4 bits or 5 bits may be used.

Further, the adaptive quantization is not limited to binarization, but may be ternarization.

Also, the coding system is not limited to the above-mentioned embodiment, and MH coding or the like may be used for a binary image, and fractal or vector may be used for coding a multi-valued image. Quantization or the like may be adopted.

Further, although the method of using the 3 × 3 filter has been described as the edge emphasis processing, the configuration of the filter is not limited to this, and may be 5 × 5 or more.
A method that does not depend on a filter, for example, a method such as pattern matching may be used.

According to each of the above-described embodiments, an image is input, the background state of the input image is determined, optimal quantization is performed from the input image according to the background state, the edge of the input image is corrected, and the edge is corrected. Optimal quantization is performed from the emphasized image according to the background state, the optimal quantization result for region determination is generated by integrating these quantization results, and the region determination is performed based on the integrated quantization result. Thus, by accurately switching the plurality of coding means for performing the optimum coding according to the characteristics of the image in accordance with the result of the area determination means, it is possible to perform image coding with high image quality and high coding efficiency.

Further, by emphasizing the edges, it is possible to improve the accuracy of the determination of the character area including many edge portions with a simple structure.

Further, it is preferable to extract whether or not the background can be efficiently coded by the binary image coding by extracting the pixels forming the flatness from the image and judging the state of the background of the image from the frequency distribution thereof. Can be determined.

Further, by determining the background of the image from the extracted frequency distribution of the background pixels, the image can be viewed as a whole, and the area in which the binary image encoding can be performed can be suitably determined, and at the same time, the motion can be determined. In the arithmetic arithmetic coding, the area is increased, so that a better convergence rate can be obtained.

In addition, in the case where the background of the image is plain, in order to extract the background density and appropriately perform the area determination,
Since quantization is performed in consideration of the background density, it is possible to provide information useful for image area separation by performing area determination based on the quantization result.

When the background of the image is a pattern, it can be quantized so that the area can be determined.
Even when the pixel value distribution is as shown in FIGS. 8A and 8B, the area determination can be performed.

Further, by using the smoothing means, it is possible to remove noise and efficiently extract the pixels forming the flatness from the image.

By repeating the region division, the region can be subdivided, the coding efficiency can be further enhanced, and the image quality can be improved.

As described above, it is possible to suppress the deterioration of the high frequency area in the character or line drawing area and improve the quality of the decoded image.

The present invention may be applied to a system composed of a plurality of devices or an apparatus composed of a single device.

It goes without saying that the present invention can also be applied to the case where it is achieved by supplying a program to a system or an apparatus.

[0118]

As described above, according to the present invention,
For example, it is possible to provide an image processing apparatus and a method for correctly dividing an input image into regions based on characteristics and performing encoding according to the characteristics of the divided areas.

[Brief description of drawings]

FIG. 1 is a block diagram showing a configuration example of an image encoding device,

FIG. 2 is a block diagram showing a configuration example of an image coding apparatus including an image processing apparatus according to an embodiment of the present invention,

FIG. 3 is a block diagram showing a configuration example of a background determination unit shown in FIG.

FIG. 4 is a diagram illustrating a method of extracting a maximum frequency density range;

5 is a block diagram showing a configuration example of an adaptive quantizer shown in FIG.

FIG. 6 is a block diagram showing a configuration example of an image encoding device including an image processing device according to a second embodiment of the present invention,

7 is a block diagram showing a configuration example of a background determination unit shown in FIG.

FIG. 8A is a diagram explaining a method of extracting a maximum frequency density range;

FIG. 8B is a diagram illustrating a method of extracting the maximum frequency density range;

9 is a block diagram showing a configuration example of an adaptive quantization unit shown in FIG.

10 is a block diagram showing a configuration example of the binary image encoder shown in FIG. 6,

FIG. 11 is a block diagram showing a configuration example of an image encoding device equipped with the image processing device of the third embodiment according to the present invention;

FIG. 12 is a flowchart showing an example of the encoding process of the third embodiment,

FIG. 13 is a flowchart showing an example of the encoding process of the third embodiment,

FIG. 14 is a flowchart showing an example of the encoding process of the third embodiment,

FIG. 15 is a flowchart showing an example of the encoding process of the third embodiment,

FIG. 16A is a diagram showing an example of the contents of an area information table;

FIG. 16B is a diagram showing an example of the contents of the area information table;

FIG. 16C is a diagram showing an example of the contents of the area information table;

FIG. 16D is a diagram showing an example of the contents of an area information table.

[Explanation of symbols]

 1 scanner 2 storage device 3 CPU 5 background judgment unit 6 adaptive quantization unit 7 area divider 8 area information encoder 9 multi-valued image encoder 10 binary image encoder 11 combiner 12 communication interface (communication I / F) 13 Communication line 14 Area division memory

Claims (15)

[Claims] 1. A determination means for determining a background state of an input image, a correction means for correcting the input image, and a quantization means for correcting the input image based on a determination result of the determination means. The corrected image is quantized, and a quantizing means for integrating the two quantization results, a dividing means for dividing the input image into regions based on the integrated quantization result, and An image processing apparatus comprising: an encoding unit that performs encoding according to a characteristic of a region. 2. A determination means for determining a background state of an input image, a correction means for correcting the input image, and a first quantization means for quantizing the input image based on a determination result of the determination means. A second quantizing means for quantizing the image corrected by the correcting means based on the determination result, based on the determination result, the input image and the image corrected by the correcting means, Third quantizing means for quantizing the input image; integrating means for integrating the respective quantization results obtained by the first to third quantizing means; and the input based on the integrated quantizing result An image processing apparatus comprising: a dividing unit that divides an image into regions, and an encoding unit that encodes each divided region according to the characteristics of the region. 3. The determining means includes: an extracting means for extracting pixels forming a flatness from an input image; a counting means for counting the frequency of each value of the extracted flat pixels; and a counting means for obtaining the frequency. 3. The image processing apparatus according to claim 1, further comprising a detection unit that detects a background state of the input image from the frequency distribution. 4. The determining means includes a smoothing means for smoothing an input image, an extracting means for extracting pixels forming a flatness from the smoothed image, and a smoothing means for extracting values of the extracted flat pixels. The image processing according to claim 1 or 2, further comprising: counting means for counting the frequency, and detection means for detecting the background state of the input image from the frequency distribution obtained by the counting means. apparatus. 5. The detecting means detects a first density range of the background from the frequency distribution, calculates an occupation rate of the background in the entire image based on the detected first density range, 5. The image processing apparatus according to claim 3 or 4, wherein the background state is detected based on the density width and the occupation ratio. 6. The image processing apparatus according to claim 5, wherein the detection unit further detects a second density width of a background narrower than the first density width from the frequency distribution. 7. The image processing apparatus according to claim 1, wherein the correction unit performs edge enhancement processing on the input image. 8. The quantizing means quantizes the input image and the corrected image based on the density range of the background determined by the determining means to form the quantized image of the background. The image processing apparatus according to claim 1, wherein 9. The first and second quantizing means quantize the input image and the corrected image based on the density width of the background determined by the determining means, and quantize the background. 3. The image processing apparatus according to claim 2, which forms an image. 10. The first and second quantizing means are:
The first density range or the second density range is selected according to the background state, and the input image and the corrected image are quantized based on the selected density range to form the quantized image of the background. The image processing device according to claim 6, wherein 11. The third quantizing means quantizes the input image based on the density of the background determined by the determining means, the input image and the corrected image, and quantizes the background. The image processing apparatus according to claim 2, wherein the image processing apparatus forms a digitized image. 12. The third quantizing means sets the pixel of interest to the background when the value of the pixel of interest of the input image and the value of a correction pixel obtained by correcting the pixel of interest sandwich the density of the background. 12. The image processing device according to claim 11, wherein the image processing device is quantized as pixels forming 13. The image processing apparatus according to claim 1, wherein the dividing unit performs repeated area division on an area having a characteristic of a multi-valued image. 14. A determination step of determining a background state of an input image, a correction step of correcting the input image, and a first quantization step of quantizing the input image based on a determination result of the determination step. A second quantization step of quantizing the image corrected in the correction step based on the determination result, and an integration step of integrating the quantization results obtained in the first and second quantization steps An image processing method comprising: a dividing step of dividing the input image into areas based on the integrated quantization result; and an encoding step of encoding each divided area according to the characteristics of the area. . 15. A determination step of determining a background state of an input image, a correction step of correcting the input image, and a first quantization step of quantizing the input image based on a determination result of the determination step. A second quantization step of quantizing the image corrected in the correction step based on the determination result, and the input image based on the determination result, the input image and the image corrected in the correction step A third quantization step of quantizing the input image, an integration step of integrating the quantization results obtained in the first to third quantization steps, and a region of the input image based on the integrated quantization result. An image processing method comprising: a dividing step of dividing, and an encoding step of encoding each divided area according to the characteristics of the area.