JP2001184511A - Device, system and method for processing image and storage medium - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an image processor capable of performing the optimal and efficient area division of a multilevel image. SOLUTION: A binarizing means 105 acquires plural binary images from the multilevel image. An area extracting means 107 extracts an area (black area), which includes a pixel group (the set of the groups of black pixels) having a prescribed value, from the plural binary images provided by the binarizing means 105. An area dividing means 110 divides the area provided by the area extracting means 107 on the basis of the crowding state of starting and ending pixels of a pixel group existent inside the relevant area. An area analyzing means 109 analyzes an attribute concerning all the areas provided by the area extracting means 107 and the area dividing means 110.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to, for example, an image processing apparatus, an image processing system, an image processing method, and a processing step for performing a region division of a multivalued image obtained by reading a color document. In a computer-readable storage medium.

[0002]

2. Description of the Related Art In recent years, with the spread of scanners, documents have been digitized. Thus, for example, when a document is read by a scanner and area division is performed on the read image, there are the following merits.

For example, a scanner converts a document A4 size color document including a character area portion and an image area portion into a 300d document.
When reading in full color 24 bits (multi-valued) with pi, the information amount of the read image (digitized document) is about 24 Mbytes, but the read image is divided into a character area portion and an image area portion, By performing binarization and MMR encoding for the character area portion and performing JPEG encoding and the like for the image area portion, it is possible to reduce the amount of information while maintaining the image quality of the read image. For example,
Depending on the image, the information amount can be reduced to about 1/100. As for the character area, O
If the encoding is performed by performing the CR, the file can be searched later, and the file can be reused in another text editing type application.

[0004]

However, in the conventional image area dividing method, when multi-valued images such as a read image of a color document obtained by a scanner or the like are divided, one multi-valued image is divided into one. Binarization was performed using a threshold, and region division was performed on the binarized image. For this reason, if the accuracy of the binarization result is poor, it will affect the region division,
It has been difficult to perform accurate region division. Further, in a color document, it is considered that various combinations of color portions, such as a background and a character region portion, may exist in one document. Therefore, good region division cannot be performed from one binarized image. .

[0005] Specifically, for example, it is assumed that there is a document 900 as shown in FIG. In the document 900, on the area B of the color b, the character A of "
In the area D of the color d, "Kakikukeko"
Is written in color c, and these letters A,
The brightness levels of the area B, the character C, and the area D are as shown in FIG.
As shown in (b), character A, region B, character C, region D
Are distributed in the direction of the high luminance level in this order.

When such a document 900 is read by a scanner and the read image is binarized by a certain threshold, the binarization result is obtained by the threshold (binarization threshold) shown in FIG.
9 (c) to 9 (e).

FIG. 19 (c) shows that the binarization threshold is a character A
This figure shows a binarized image 901 obtained when a value between the luminance level of the character C and the luminance level of the character C is indicated (indicated by "" in FIG. 3B). FIG. 19D shows a binarized image obtained when the binarization threshold is set to a value between the luminance level of the character C and the luminance level of the area B (indicated by “” in FIG. 19B). 902. FIG. 19 above
(E) shows a binarized image 903 obtained when the binarization threshold is a value between the luminance level of the region B and the luminance level of the region D (indicated by “” in FIG. It is a thing. Therefore, unless the binarized image 902 shown in FIG. 19D is obtained, in which the value indicated by "" in FIG. 19B is obtained as the binarization threshold, the region division in which all characters can be detected cannot be expected. .

However, if the luminance level of the character C and the luminance level of the area B are such that the luminance level of the character C <the luminance level of the area B, as shown in FIG. An optimal binarized image for division cannot be obtained.

In order to solve the above problem, there is an adaptive threshold binarization method in which the binarization is changed by changing the binarization threshold depending on the area. For example, FIG.
For a document 904 in which the character C of “Kakikokeko” is written in a whitish color c ′ (high brightness) on an area D of a dark color d ′ (low brightness) as shown in FIG. Cannot divide. That is, the luminance levels of the character A, the region B, the character C, and the region D are as shown in FIG.
Characters A, D, B, and C are distributed in the direction of the high luminance level in this order. In this state, the binarization threshold is set to a value between the luminance level of the character A and the luminance level of the region D. When (indicated by “” in FIG. 2B) is a binarization threshold, a binarized image 905 shown in FIG. 2C is obtained, and the binarization threshold is set to the luminance level of the area D and the area B. (Indicated by “” in FIG. 2B), a binarized image 906 shown in FIG. 2D is obtained, and the binarization threshold is set to the luminance level of the area B. A value between the luminance levels of the character C ((b) in FIG.
In the case of “medium”, the binarized image 907 shown in FIG. 11E is obtained, and even with the adaptive threshold binarization method, the binarized image optimal for the region division is obtained. Cannot be obtained.

In the processing of the divided areas (binary processing, etc.), a circumscribed rectangle is used for convenience, and processing is performed in the shape of the circumscribed rectangle. Therefore, for example, as shown in FIG. 22, a region 910 where white characters “ABCDEFG” and “Irohanihohet” exist on a black background and “1” on a white background
23 "," 456 "," 798 ", and other areas in which black characters are present.
Is an image existing in a shape circumscribing the region 910 (shape of a circumscribed rectangle), and performing a binarization process for OCR in units of the circumscribed rectangle, as shown in FIG. So, "123", "456",
An area 911 where a black character such as “789” exists on a white background
Is inverted in spite of being a normal area (an area where a black character exists on a white background).

If the binarization process at this time is not a process in units of circumscribed rectangles but a process in free form, the above problem does not occur, but the process in free format supports it. Functions and the like are required, which complicates the processing. On the other hand, the processing in units of the circumscribed rectangle is
Although the processing can be efficiently performed with a simple configuration, the above problem occurs.

SUMMARY OF THE INVENTION The present invention has been made to eliminate the above-mentioned disadvantages, and has been made in view of the above circumstances.
It is an object of the present invention to provide an image processing method and a storage medium in which processing steps for executing the image processing method are readable by a computer.

[0013]

For such a purpose,
According to a first aspect, there is provided an image processing apparatus for performing region division of a multi-valued image, comprising: a quantization unit configured to acquire a plurality of quantized images from the multi-valued image; A region extraction unit that extracts a region including a pixel group having a predetermined value from an image, and a region obtained by the region extraction unit is determined based on a density of start pixels and end pixels of the pixel group existing in the region. And a region analyzing unit that analyzes attributes of each region obtained by the region extracting unit and the region dividing unit.

According to a second aspect, in the first aspect,
The quantizing means acquires a binarized image as the quantized image.

According to a third aspect, in the first aspect,
The quantizing means acquires a quantized image of an area portion of the multi-valued image based on an area extraction result of the area extracting means.

According to a fourth aspect, in the first aspect,
The quantizing means acquires the plurality of quantized images based on a plurality of thresholds obtained from a histogram of the multi-valued image.

According to a fifth aspect, in the first aspect,
The region extracting unit includes a detecting unit configured to detect a pixel group in which the pixels having the predetermined value are one-dimensionally continuous from the quantized image, and a second unit that stores coordinate information of the pixel group detected by the detecting unit. (1) holding means, second holding means for holding the coordinate information of the pixel group already detected by the detection means, and combining information in the first holding means and information in the second holding means. And coupling means for performing processing.

According to a sixth aspect based on the fifth aspect,
The detection means detects the area on the quantized image for every M lines.

According to a seventh aspect, in the first aspect,
The region extracted by the region extracting unit includes a circumscribed rectangular region including a one-dimensionally continuous pixel group, and the region dividing unit includes a pixel included in the region extracted by the region extracting unit. A first method for detecting a dense point of the start pixel and the end pixel from the histogram of the start pixel and the end pixel of the group.
Detecting means for detecting a spread in a dimensional direction different from the one-dimensional direction with respect to a start pixel and an end pixel of a pixel group present in a divided area obtained based on a detection result obtained by the first detection means. 2 and a dividing means for performing region division based on the detection result of the second detecting means.

According to an eighth aspect of the present invention, there is provided an image processing system in which a plurality of devices are communicably connected, wherein at least one device among the plurality of devices is according to any one of claims 1 to 7. It has a function of an image processing device.

A ninth aspect of the present invention is an image processing method for dividing a multi-valued image into regions, wherein at least one or more quantized images are created from the multi-valued image, and An area extracting step of obtaining an area of a cluster of black pixels, and dividing the black area obtained by the area extracting step on the basis of the density of the start pixels and end pixels of the cluster of black pixels existing in the black area. And a region analyzing step of analyzing attributes of all regions obtained by the region extracting step and the region dividing step.

In a tenth aspect based on the ninth aspect, the area extracting step includes a step of creating a binarized image as the quantized image.

In an eleventh aspect based on the ninth aspect, a part of the at least one or more quantized images is not a quantized image obtained from the entire surface of the multi-valued image. .

In a twelfth aspect based on the ninth aspect, the region extracting step includes a black detecting step of detecting a cluster of one-dimensionally continuous black pixels from the quantized image, and a black detecting step. A first holding step of holding the detected coordinate information of the cluster of black pixels, a second holding step of holding the coordinate information of the cluster of black pixels already detected by the black detection step, and the first holding step; The method includes a combining step of combining the held information in the holding step and the held information in the second holding step.

In a thirteenth aspect based on the twelfth aspect, the black detecting step includes a step of executing the detecting process for each M lines on the quantized image.

In a fourteenth aspect based on the ninth aspect, the black area extracted in the area extracting step is:
A circumscribed rectangular area including a cluster of one-dimensionally continuous black pixels is included, and the area dividing step includes a start pixel and an end pixel of a cluster of black pixels included in the black area extracted by the area extracting step. A first detection step of detecting a dense point of the start pixel and the end pixel from the histogram, and dividing the black region based on a detection result in the first detection step; A second detection step of detecting a spread in a dimensional direction different from the one-dimensional direction for the start pixel and the end pixel of the group, and dividing the black area based on a detection result in the second detection step. And a dividing step to be executed.

According to a fifteenth aspect, a processing program for implementing the function of the image processing apparatus according to any one of claims 1 to 7 or the function of the image processing system according to claim 8 is provided.
It is a storage medium that is readable by a computer.

According to a sixteenth aspect of the present invention, a computer readable storage medium stores the processing steps of the image processing method according to any one of the ninth to fourteenth aspects.

[0029]

Embodiments of the present invention will be described below with reference to the drawings.

The present invention is applied to, for example, an image processing apparatus 100 as shown in FIG. This image processing apparatus 100
As shown in FIG. 1, the luminance conversion unit 10 acquires luminance information (luminance value) from input image information (multi-valued image information).
1, a multi-valued image memory 102 for holding the luminance information obtained by the luminance converting unit 101 for one image, a stogram calculating unit 103 for generating a histogram based on the luminance information in the multi-valued image memory 102, and a histogram calculating unit A threshold determination unit 104 for determining a plurality of thresholds based on the histogram obtained in 103; an area analysis unit 109 for obtaining an optimal binarization threshold or the like based on the histogram obtained in the histogram calculation unit 103; A binarizing unit 105 for generating a binarized image based on a threshold value obtained by the unit 104 or the region analyzing unit 109; and a binary image memory for holding the binarized image obtained by the binarizing unit 105 106, a black region extracting unit 107 for obtaining coordinate information of the black region from the binary image in the binary image memory 106, and analyzing the black region obtained by the black region extracting unit 107 Obtaining information of a plurality of black regions by dividing the black area as needed Te black area dividing unit 110
And a black region coordinate storage unit 108 for storing information obtained by the black region division unit 110 and the region analysis unit 109. The histogram calculation unit 103 and the binarization unit 105
Are executed based on the information in the black area coordinate storage unit 108.

The image processing apparatus 100 operates as follows according to, for example, the flowchart of FIG.

Step S201: The multi-valued image information (hereinafter referred to as "target image") input to the luminance conversion unit 101 is color image information composed of information of three primary colors of RGB. The luminance conversion unit 101 generates luminance information (luminance value Y) for each pixel constituting the target image input in RGB.
Is calculated according to the following equation: Y = 0.299R + 0.587G + 0.114B. The multi-valued image memory 102 holds the luminance values obtained by the luminance conversion unit 101 for the pixels constituting one image (target image).

Step S202: Histogram calculation section 1
In step 03, the luminance values Y of the pixels constituting the entire image or a partial image are read from the multi-value image memory 102, and a histogram of the luminance is generated. Here, a histogram of the luminance values of all the pixels constituting the target image held in the multi-valued image memory 102 is generated. At this time, since the total number of pixels changes depending on the resolution and image size at the time of image reading, it is preferable to perform normalization based on the resolution and image size at the time of image reading. This is achieved by using the state of transition of the histogram obtained by the histogram calculation unit 103 when determining a plurality of threshold values in the threshold value determination unit 104 described later, and determining the inclination of the change in the number of pixels and the number of pixels themselves. This is because it is a parameter.

Step S203: The threshold value determining unit 104
A plurality of thresholds are determined based on the histogram obtained by the histogram calculator 103.

Specifically, for example, as an object image, FIG.
When multivalued image information of a document as shown in FIG. 9A is input and the luminance information is held in the image memory 102,
The histogram obtained by the histogram calculator 103 is a histogram as shown in FIG.

Therefore, the threshold value determining unit 104 first determines the luminance value “255” (white) in the histogram of FIG.
From the pixel to the brightness value “0” (black). At this time, the reference points are sampled at intervals of four (“255”, “251”, “247”,...).
・), Etc.

Then, the threshold value determination unit 104 detects a luminance value corresponding to “satisfies condition 1> satisfies condition 2> satisfies condition 3” in the following conditions 1 to 6 as a threshold value. . Condition 1: The total number of pixels from the reference point to the point that goes up by "10" is larger than the threshold value 1. Condition 2: A sharp decrease has occurred (“greater than slope 1”)
Once, or the state of “greater than slope 2” occurs twice in succession). Condition 3: A gradual decrease or increase has occurred (slope 3
Smaller state). Condition 4: The total number of pixels from the reference point to a point that is “40” up is larger than the threshold value 2. Condition 5: the number of pixels at the reference point is smaller than threshold value 3. Condition 6: the number of pixels whose luminance value is “20” before the reference point is larger than threshold value 4.

Therefore, as shown in FIG.
Four thresholds indicated by 01 "to" 404 "are detected. Among these thresholds 401 to 404, for example, regarding the threshold 402," condition 4 is satisfied, condition 5 is satisfied, and condition 6 is satisfied. It is detected as a threshold value corresponding to “satisfied”. That is, the conditions 1 to 3 are conditions for extracting a cluster of luminance with little variation, and the conditions 4 to 6 are conditions for extracting a cluster of luminance that is slowly distributed. The threshold 402 is detected as a threshold corresponding to the conditions 4 to 6 among these conditions.

Thereafter, as a post-processing, the threshold determination unit 104 refers to the number of pixels corresponding to the luminance value “0” to the luminance value “255” in the histogram of FIG. The total value is calculated, and the threshold values 401 to 4
Among 04, the threshold value corresponding to the luminance value existing where the cumulative number of pixels is smaller than the threshold value 5 is deleted. As a result, the threshold 404 among the thresholds 401 to 404 is deleted. The threshold determination unit 104 determines the remaining thresholds 401 to 403
Is determined as a binarization threshold value to be given to the binarization unit 105.

In the above-described threshold determination section 104, the erasure processing of the threshold in the post-processing is not an essential processing.
Also, in the conditions 1 and 4, the value going up from the reference point is not limited to “10” or “40”, but may be any value (a value depending on the characteristics of the scanner). You may. Further, the processing in the threshold value determination unit 104 may be executed while averaging so as not to be confused by noise on the histogram as shown in FIG. Further, when detecting the threshold value according to the conditions 1 to 6, if the conditions 1 to 6 occur once at the reference point, the luminance value of the reference point is not determined as the threshold value, but, for example, two consecutive times. When the occurrence occurs, the luminance value of the reference point may be determined as the threshold.

Step S204: The binarization section (simple binarization section) 105 converts the plurality of binarization threshold values obtained by the threshold value determination section 104 or the threshold values described later obtained by the region analysis section 109. Based on the target image (the luminance information of each pixel forming the target image)
Is binarized to generate a binarized image. Binary memory 10
Reference numeral 6 holds the binarized image obtained by the binarization unit 105.

Therefore, for example, the multi-valued image information of the document shown in FIG. 3A is input as the target image, and the binarization unit 105 outputs a plurality of binarized images obtained by the threshold value determination unit 104. When a binarized image is generated based on the thresholds 401 to 403, the binarized image as shown in FIGS. 3B to 3C is held in the binary memory 106. Become. FIG. 3B shows a binarized image obtained by binarization based on the binarization threshold 401, and FIG. 3C shows a binarized image based on the binarization threshold 402. FIG. 3D shows a binarized image obtained by binarization, and FIG. 4D shows a binarized image obtained by binarization based on the binarization threshold 403.

Step S205: Black area extraction unit 107
Creates coordinate information of a black area from the binarized image held in the binary memory 106. For example, although the details will be described later, the black area extraction unit 107 selects “white” (luminance value “255”) among the binarization thresholds 401 to 403 shown in FIG.
The binarized image (see FIG. 3B) obtained by the binarization based on the threshold value 401 closest to the luminance value corresponding to is read out from the binary image memory 206, and in the binary image, By extracting the coordinates of a cluster of pixels having a pixel value corresponding to “black”, coordinate information of a black area is created. Similarly, coordinate information of a black region is created for a binarized image obtained by another threshold. The black region dividing unit 110 analyzes the black region obtained by the black region extracting unit 107, and further divides the black region as necessary to obtain information on a plurality of black regions. Black area coordinate holding unit 10
Numeral 8 holds the coordinate information of the black area obtained by the black area dividing unit 110.

Step S206: The area analysis unit 109
As will be described in detail later, the coordinate information of all the black areas held in the black area coordinate holding unit 108 is compared with the image information (luminance information held in the multi-valued image memory 102) corresponding to the target black area. The histogram is calculated by the histogram calculation unit 10
3 to determine whether or not the target black area is a character area, and if it is a character area, whether or not the target black area is an inverted area (an area where a white character exists on a black background), Analysis information (analysis information of the attribute of the black area) such as a pixel value (average background color), an average pixel value (average character color) of the character, and an optimal binarization threshold used for binarization are acquired. The black area coordinate storage unit 108 stores various analysis information obtained by the area analysis unit 109.

The processing indicated by the dotted arrow shown in FIG. 2 will be described later. Further, in the present embodiment, RGB color image information is input. However, the present invention is not limited to this, and any color space information such as CMYK or Lab may be used. In addition, RGB color image information is converted into luminance information and luminance is one-dimensionally binarized. However, the present invention is not limited to this.
Binarization may be performed using an RGB three-dimensional space.

FIG. 5 is a flowchart showing the operation of step S20 shown in FIG.
5 specifically illustrates the black region extraction processing of FIG. That is, the black region extraction unit 107 operates as follows according to the flowchart of FIG.

Step S501: First, for example, among the binarization thresholds 401 to 403 shown in FIG.
(Binarized image obtained by binarization based on the binarization threshold 401 closest to the luminance value corresponding to (luminance value “255”) (see FIG. 3B, hereinafter, “target binary”). ) Is read out from the binary image memory 206, and a counter j () for executing the present process for the target binary image for each number of lines indicated by a preset sampling parameter sample_state The counter in the vertical direction (sub-scanning direction) is reset to “0”.

Step S502: It is determined whether or not the counter j has reached a preset value height (the vertical width of the target binarized image). As a result of this determination, "j <hei
If it is not “ght”, it recognizes that this processing for the target binary image has been completed, and ends this processing. On the other hand, “j <h
If "right", the processing from the next step S503 is executed.

Step S503: If "j <height" is determined as a result of the determination in step S502, as will be described in detail later, in each pixel constituting one line on the target binary image indicated by the counter j, " A set of pixels (black pixels) having a pixel value corresponding to “black” is detected. The number of sets of black pixel clusters detected in step S503 is indicated by “linenum”. In a set of one black pixel block, the coordinates of the start pixel of the block are indicated by xstart [x], and the coordinates of the end pixel are indicated by xend [x].

Step S504: Each set ("linenum") of a set of black pixels detected in step S503
Step S)
Counter l for executing the processing of 505 to S509
Is reset to “0”.

Step S505: Counter l is set to l
It is determined whether or not inenum has been reached. As a result of this determination, if “l <linenum” is not satisfied, “linen
um ”for all sets of several minutes.
It recognizes that the processes of 505 to S509 have been executed, and proceeds to step S510 described later. On the other hand, "l <line
num ”, the processing from the next step S506 is executed.

Step S506: If "l <linenum" as a result of the determination in step S505, the coordinates xstart of the set of black pixel clusters indicated by the counter l
[L] and the coordinates xend [l] are compared with the coordinates of all the existing structures Group described later, and the coordinates x
A connection process is performed between a group of black pixels indicated by start [l] and coordinates xend [l] and the structure Group.

Step S507: Coordinate xstart
It is determined whether or not the cluster of black pixels indicated by [l] and the coordinate xend [l] has not been combined with any of the above-mentioned existing structures Group. As a result of this determination, the coordinates xstart [l] and the coordinates xend
If the cluster of black pixels indicated by [1] is combined with any of the existing structural groups, the next step S508 is not executed, and the process directly proceeds to step S509 described later.

Step S508: As a result of the judgment in step S507, the coordinates xstart [l] and the coordinates xend
If the cluster of black pixels indicated by [l] is not connected to any of the above-mentioned existing structures Group, the coordinates xstart [l] and the coordinates xend
The cluster of black pixels indicated by [l] is registered in a new structure Group. This registered structure Grou
p is used in the combining process in step S506 described above. The details of the “structure group” here will be described later.

Step S509: coordinate xstart
The counter l is incremented (l = l + 1) to execute the processing from step S505 on the next set of black pixels indicated by [l] and the coordinates xend [l]. It returns to S505.

Step S510: As described above, steps S505 to S505 are executed for all the sets of the black pixel clusters.
When the processing of step S509 is completed, the sampling parameter sampl is added to the vertical counter j.
e_state is added, and the process returns to step S502.
Thereby, the sampling parameter sample_t
If ate is, for example, “1”, this processing is executed for each line of the target binary image, and if “ate” is “4”, 4 is applied to the target binary image. This processing is executed for each line.

FIG. 6 is a flowchart showing step S50 shown in FIG.
3 specifically illustrates a detection process of a set of a set of black pixels of No. 3;

Step S601: First, for every pixel constituting the target line (the line in the target binary image indicated by the vertical counter j), for each pixel number indicated by the preset sampling parameter sample_yoko Then, the counter i (the counter in the horizontal direction (main scanning direction)) for executing this processing is reset to “0”. In addition, a flag "startflag" used in the step processing described later is set to "OFF", and "linenum" indicating the number of groups of black pixels is set.
Is reset to “0”.

Step S602: It is determined whether or not the counter i has reached a predetermined value width (width of the target binarized image). As a result of this determination, “i <widt
h ”, the process proceeds to step S613 described below,
If “i <width”, the next step S603
Proceed to.

Step S603: If “i <width” as a result of the discrimination in step S602, a window area of width m × n pixels is provided around the pixel indicated by the counter i in the target line, and the area is set. , Pattern matching is performed using preset pattern data of horizontal m × vertical n pixels. Then, it is determined whether or not the window area matches the pattern data. As a result of this determination, if they match, the process proceeds to the next step S604, and if they do not match, the process proceeds to step S607 described later.

In step S603, the window area of horizontal m × vertical n pixels for performing pattern matching may be, for example, an area which is referred to only in the horizontal direction, such as horizontal 7 × vertical 1 pixel, or horizontal. It may be an area that is vertically referenced, such as 3 × 3 vertical pixels. In addition, instead of pattern matching, the process proceeds to step S604 using conditions such as the following conditional expressions 1 and 2, or proceeds to step S607.
It may be determined whether to proceed to. Conditional expression 1: The pixel indicated by the counter i (pixel of interest) is a black pixel. Conditional expression 2: Two pixels adjacent to the target pixel are black pixels.

Step S604: If the window area and the pattern data match as a result of the determination in step S603, the flag startflag is set to "OF".
F "is determined. As a result of this determination," st "
If "artflag = OFF", the next step S
Proceeding to step 605, if not “startflag = OFF”, the next step S605 and S606 are not executed, and the process directly proceeds to step S612 described later.

Step S605: If “startflag = OFF” as a result of the determination in step S604, xs indicating the coordinates of the first pixel of the set of black pixels is set.
A counter i is set for start [linenum]. That is, with respect to xstart [linenum] indicating the coordinates of the first pixel of the set of black pixels (target set) indicated by the current value of “linenum”, the coordinates of the pixel (target pixel) indicated by the counter i Set. Also, for the flag startflag, xst
art [linenum] has been set and xend
“ON” is set to indicate that the setting to [linenum] has not been completed.

Step S612: The sampling parameter sample_yoko is added to the pixel of the counter i (counter in the horizontal direction (main scanning direction)), and the process returns to step S602. Accordingly, when the sampling parameter sample_yoko is, for example, “1”, the present process is executed for each pixel on the target line, and when the sampling parameter sample_yoko is “4”, 4 for the target line.
This processing is executed for each pixel. If it is determined in step S602 that “i <width” is not satisfied, the process proceeds to the next step S613.

Step S613: If the result of determination in step S602 is that "i <width" is not satisfied, the flag st
It is determined whether or not artflag is “ON”. If the result of this determination is not “startflag = ON”, that is, not “i <width” and “staflag = ON”,
If “rtflag = ON” is not satisfied, it is recognized that the detection of all the sets of black pixels on the target line has been completed, and this processing ends. On the other hand, if “startflag = ON”, the process proceeds to step S608 described below.

Step S607: Also, in the above-mentioned step S603, a window area of horizontal m × vertical n pixels centering on the pixel indicated by the counter i of the target line and a predetermined pattern of horizontal m × vertical n pixels If it is determined that the data does not match, the flag startfl
It is determined whether or not ag is “ON”. If the result of this determination is not “startflag = ON”, the processing proceeds to the above-described step S612 without executing the processing of the next steps S608 to S611. on the other hand,
If “startflag = ON”, the process proceeds to the next step S608.

Step S608: In step S613 or step S607, "startflag = ON"
When it is determined that “xend [linenum]” indicates the coordinates of the pixel at the end of the set of black pixels, “counter i−1” is set. That is, "li
set of black pixels indicated by “nenum” (target set)
Xend [linemenu indicating the coordinates of the pixel at the end of
m], the pixel indicated by the counter i (pixel of interest)
Set the coordinates of the pixel before.

Step S609: "xend [line
It is determined whether or not the result of [num] -xstart [linenum] +1 "is greater than a preset limit value del_width of the target binarized image. As a result of this determination,""xend [linenum]- xst
art [linenum] +1 "> del_widt
If it is not "h", the process of the next step S610 is not executed, and the process directly proceeds to step S611, which will be described later.
xend [linenum] -xstart [line
num] +1 ">del_width",
The process proceeds to the next step S611.

Step S610: As a result of the determination in step S609, "" xend [linenum] -xsta
rt [linenum] +1 ">del_width"
, The set of pixels indicated by “xstart [linenum]” and “xend [linenum]” is determined as the set of black pixels indicated by the current value of “linenum”. "Linenum" is incremented for the detection of.

Step S611: Flag startfl
ag is set to “OFF” and the above-described step S61 is performed.
Proceed to 2. Note that as a result of the determination in step S609, "" x
end [linenum] -xstart [linen
um] +1 ”> del_width”, the step S609 is not executed, ie, “line
num "is not incremented, so the current" xs
start [linenum] "and" xend [lin
enum] ”is overwritten in the processing from step S611. That is,“ xstart [linen]
um] "and" xend [linenum] "are not recognized as a set of black pixels, and are effectively deleted.

FIG. 7 shows an example of a structure of a set of a group of black pixels obtained as described above.

In the processing shown in FIG. 6, a set of black pixels is converted into a structure Part 701 indicated by "701" in FIG.
s. Here, the window to be opened is one-dimensional and has only the coordinates in the X direction. However, the present invention is not limited to this.
As shown by "a", the window may be two-dimensional, and the members of the structure Parts may include information in the y direction.

The structure Parts is the xstar described above.
t [linenum] and xend [linenum]
(Members) "startx" and "end" indicating the start coordinates and end coordinates of a set of black pixels corresponding to
x "and address information" next "of the next information, and" NULL "is set in the address information" next "when there is no next information.

In the processing shown in FIG. 5, a set of black pixels (structure Pa) obtained by the processing shown in FIG.
rts) are combined, and a structure Group indicated by “702” in FIG. 7 is generated. The structure Group includes information “sx” and “sx” that indicate the start coordinate and the end coordinate two-dimensionally.
sy ”,“ ex ”, and“ ey ”, and the structure Group
The information “first” of the address of the first structure Parts of the collection of the structures Parts that constructs the structure “Parts”, and the information “last” of the address of the last structure Parts of the collection of the structures Parts that construct the structure Group These pieces of information are generated by the black region extraction unit 107.

The structure Group is composed of the structure Gro
Threshold information “Threshold” suitable for binarizing the area indicated by “up”, and flag information “CharaFlg” indicating whether the area indicated by the structure Group is a character area or a non-character area (eg, an image area). And flag information "InvertF indicating whether or not the region indicated by the structure Group is a character region, whether or not the region is an inverted character region (a black background and a white character portion).
lg "and flag information" Shape indicating whether or not the shape of the black area portion of the area indicated by the structure Group is square.
Rectangle ”indicates whether or not substantially the same object exists on the binarized image created with the threshold value one immediately before the threshold value that created the binarized image in which the region indicated by the structure Group is extracted. Flag information "SameAsMo
ther "and pointer information" next "indicating the address of the next structure Group.
It is generated by the area analysis unit 109. Pointer information "ne
xt ”includes“ N ”when there is no next structure Group.
"ULL" is set.

FIGS. 8A to 8F show the black region extracting unit 10.
By executing the processing shown in FIGS. 5 and 6 in FIG. 7, what kind of processing is actually performed on the target binarized image is specifically shown.

Referring to FIG. 8A, first, in step S503 shown in FIG. 5, the line of the y coordinate = “5” on the binarized image (the counter j in the processing of FIG.
Xstart on the target line when it is 5 ")
A set (structure Parts) 801 of a group of black pixels indicated by [5] and xend [15], and xstart [5]
0] and xend [65], and a set (structure Parts) 802 of a group of black pixels is detected.

At this time, the existing structure Group
In step S508, two new structure groups are created, and as a member of these structure groups, the structure Parts80 is created.
1, 802 are respectively registered (set).

In this case, with respect to the structure Parts 801, the structure Parts and the structure G shown in FIG.
The following values are set as the values of the members of the group. startx = 5 (value of xstart [5]) endx = 15 (value of xend [15]) sx = 5 (value of xstart [5]) sy = 5 (value of counter j) ex = 15 (xend [15] Ey = 6 (value of “j + sample_date” where sample_state = 1) first = address of structure Parts 801 last = address of structure Parts 801 Also, for structure Parts 802, structure Pa
As in the case of rts801, values corresponding to the members of the structure Parts and the structure Group shown in FIG. 7 are set. In FIG. 8, the structure Pa
rts is shown in black, and the structure Group is shown in a white frame.

Referring to FIG. 8B: Next, at step S5
At 03, the line next to the line of the y coordinate = “5” on the binarized image (counter j is counted up to “6”)
, A set (structure Parts) 803 of black pixels indicated by xstart [0] and xend [10], and black pixels indicated by xstart [45] and xend [75] It is assumed that three sets of a set (structure Parts) 804 and a set of black pixels (structure Parts) 805 indicated by xstart [100] and xend [110] are detected.

At this time, the already existing structure Group
As the structure G of the structures Parts 801 and 802
Since the groups 801 ′ and 802 ′ exist, in step S 506, the joining processing of the structural parts 803 to 805 (joining processing by coordinate comparison) to the structural groups 801 ′ and 802 ′ is performed.

By the above coupling process, for example, the structure Pa
Since the rts 803 is in contact with the structure Group 801 ′, it is coupled to the structure Group 801 ′. As a result, the structure Group 801 ′ corresponds to FIG.
The structure grows to the structure Group indicated by “806” in (c). In addition, the structure Parts 804 is a structure Grou.
Since it is in contact with p802 ', its structure Group
802 '. As a result, the structure Group8
02 ′ is the structure G shown in “807” in FIG.
grow into a loop. Since the structure Parts 805 does not come into contact with any of the structure groups 801 ′ and 802 ′ (cannot be combined), step S508 is performed.
, A new structure Group is created, and as a member of the structure Group, the structure Parts8
05 is registered (set).

Referring to FIG. 8C: Next, at step S5
At 03, the line next to the line at y coordinate = “6” on the binarized image (counter j is counted up to “7”
Xstart [65] on the target line when
Suppose that a set (structure Parts) 808 of a group of black pixels indicated by xend [105] and xend [105] is detected.

At this time, the already existing structure Group
Since there is a structure Group 806, 807 and a structure Group 805 ′ of the structure Parts 805, in Step S506, the structure Group 8
Structure Parts80 for 06,807,805 '
8 is performed (coupling processing by coordinate comparison).

Here, the structure Parts 808 is in contact with the two structures Group 807 and 805 '. Therefore, in this case, the structure Parts 808
Is bonded to the structure Group that combines the two structures Group 807 and 805 ′. As a result,
The two structures Group 807, 805 'are shown in FIG.
It grows into one structure Group shown by “809” in (d).

In this case, information of the structure Group 807 (see FIG. 7) is used as the structure Group 809, and the structure Group 805 'is deleted. However, in the information of the structure Group 807, next = “firsts” of the structure Group 805 ′
address of the structure Parts indicated by “t” last = “las” of the structure Group 805 ′
The information of “next” and “last”, which is the address of the structure Parts indicated by “t”, is rewritten, and the result is the information of the structure Group 809.

8 (d) and (f): Therefore, the structure Gr including the structures Parts 801 and 803
up806 and the structures Parts802, 804, 8
Two structures Group are generated, including a structure Group 806 including 05,808.

The black region (structure Gro) as described above
The extraction process of the region indicated by “up” is performed by the binarization unit 10.
5 is performed on each of the plurality of binarized images (binary images obtained based on the plurality of binarization thresholds).

As a result, for example, the target image is
In the case of a multi-valued image of a document as shown in (a), black regions 301 to 305 are extracted from the binarized image based on the binarization threshold 401 as shown in FIG. Is done.
Structure Group showing these black regions 301 to 305
Then, the flag information “ShapeRectangle” indicating whether or not the shape of the black area portion is a square is set to “TRU”.
E (the shape of the black area portion is a square) ", and the binary value created by the threshold value immediately before the threshold value used to create the binary image from which the area indicated by the structure Group is extracted. The flag information “SameAsMother” indicating whether or not substantially the same object exists on the structured image is “FA
LSE (the same object does not exist) "is set.

Further, as shown in FIG. 3C, the black area 30 based on the binarized image
6 to 309 are extracted. These black areas 306 to 30
9, the flag information “Shap
eRectangle "is set to" TRUE (the shape of the black area is a square). "
For “SameAsMother”, the black area 306
Is set to “TRUE (the same object exists)” and the other black area 3
For 07 to 309, "FALSE (the same object does not exist)" is set.

Further, as shown in FIG. 3D, the black area 31 based on the binary threshold
0 is extracted. The structure Gro indicating the black region 310
In the up, the flag information "ShapeRectangl
e "is set to" FALSE (the shape of the black area is not square) "and the flag information" SameAsM
"other" is set to "TRUE (the same object exists)".

When the black area is extracted, the flag information “Sh
A region in which “aperRectangle” is “TRUE” or a region having a predetermined size or more may be extracted as a black region.

The black area extraction unit 107 has, for example, a configuration as shown in FIG. 9 as a configuration for performing the above-described processing (the processing shown in FIGS. 5 and 6). That is, as shown in FIG.
A window is provided for the binarized image to be processed held in the binary memory 106 (see FIG. 2), and a set of black pixels in the window (a set of black pixels indicated by the structure Parts) ) Detecting section 100
1, a coordinate holding unit 1002 that holds one-dimensional coordinate information of a set of black pixels detected by the detection unit 1001 (coordinate information of a set of black pixels indicated by the structure Parts), A two-dimensional coordinate storage unit 1004 for storing two-dimensional coordinate information (coordinate information of a black area indicated by the structure Group) of a group of completed black pixels, and the coordinate information and two-dimensional coordinate storage in the coordinate storage unit 1002 Part 10
The two-dimensional coordinate information (structure Gro) obtained as a result of comparing the coordinate information in the data structure 04 and executing the combining process (the merging process of the structure Parts into the structure Group) as necessary
and a comparison unit 1003 for holding the coordinate information of the black area indicated by “up” in the two-dimensional coordinate holding unit 1004. The information held in the two-dimensional coordinate holding unit 1004 is stored in the black area dividing unit 110 (see FIG. 2).

The black area dividing section 110 includes the black area extracting section 10
When the information on the black area is given from 7, for example, the following operation is performed according to the flowchart shown in FIG.

Step S1101: First, all black areas (black areas indicated by the structure Group, hereinafter referred to as “structures Group”) obtained from one binarized image from the black area extraction unit 107.
It is determined whether the processing from the next step S1102 has been completed on the “structure group”). If the result of this determination is that the processing has been completed, this processing is terminated. If there is an unprocessed structure Group, the flow proceeds to the next step S1102.

Step S1102: Referring to the information of the unprocessed target structure Group (see FIG. 7), whether the size of the target structure Group is larger than a preset threshold value or the shape of the black area portion The flag information “ShapeRectangle” indicating whether or not it is a square is “FA”
LSE (the shape of the black area is not square) ”is determined, and as a result of the determination, the size of the target black area is larger than a preset threshold value, and the flag information“ Shape ”
If "Rectangle" is "FALSE", the flow advances to the next step S1103; otherwise, the flow directly advances to step S1110 to be described later.

Step S1103: Step S1102
When the target structure Group satisfies the above condition as a result of the determination in the above, the structure P included in the target structure Group
With reference to the information of the arts (see FIG. 7 above), the structure Pa
A histogram of “startx” (hereinafter, “startx”) indicating the coordinates of the first pixel of the set of black pixels indicated by rts
"End" indicating the coordinates of the pixel at the end of the cluster of black pixels.
x "histogram (hereinafter" histogram endhi
st ").

Step S1104: Step S1103
M dense points of the histogram starthist and the histogram endhist created in step (1) are detected.

Step S1105: Step S1104
With reference to the information of the crowded points obtained in the above, the target structure Gr
whether or not to split “up”, that is, the target structure Gro
It is determined whether or not the black area indicated by up should be divided. For example, an arbitrary condition is provided for the position and the number of the crowded points, and when the condition is satisfied, it is recognized that the area is divided, and the process proceeds to the next step S1106. On the other hand, when the above condition is not satisfied, it is recognized that the area is not divided, and the process proceeds to step S1110 to be described later.

Step S1106: Step S1105
When the region is divided as a result of the
One point is selected from each of the m dense points detected from the rthist and the m dense points detected from the histogram endhist, and these two points are defined as “startpoint” and “endpoint”.
In this case, the area that can be divided at a time is three areas indicated by 0 to start point start point to end point end point to the end.

Step S1107: Step S1106
Is different from the direction (x direction) of the set of black pixels indicated by the structure Parts for each of the three regions obtained by setting the division point of, for example, y
Detects directional spread.

Step S1108: Step S1107
, It is determined again whether or not to divide the area. For example, for each region of
determine whether the distribution in the y-direction is widening as a whole,
As a result of the discrimination, if the respective regions have a wide distribution in the y direction, it is recognized that the region should not be divided, and step S1110 to be described later is performed as it is.
Proceed to. On the other hand, if there is a bias in the y-direction distribution of each of the regions, it is recognized that the region should be divided,
Proceed to the next step S1109.

Step S1109: Step S1108
If the result of the determination is that the area is to be divided, step S1106
In the three areas of ~ obtained by setting the division point of
The black area indicated by the target structure Group is divided. Further, a new structure Group corresponding to the three regions is created, and the corresponding new two-dimensional coordinate information is set for the new structure Group. Then, the target structure Group that is the source of the division is deleted. The coordinate information may be updated by diverting the target structure Group from which the division is based, instead of deleting the target structure Group.

Step S1110: Step S1109
, A new structure Group obtained as a result of the division
Is stored in the black area coordinate storage unit 108 (see FIG. 2), the target structure Group is set as the next structure Group (the pointer is advanced to the next structure Group), and the process returns to step S1101 to perform the subsequent processing steps. Is repeatedly executed. On the other hand, steps S1102, S1105, S
When the step S1110 is executed in the case of branching at “No” in each determination process of 1108, the target structure Group that has not been divided is stored in the black region coordinate storage unit 108 (see FIG. 2 above) as it is. , Target structure Grou
Let p be the next structure Group (next structure Group
p), the process returns to step S1101,
The subsequent processing steps are repeatedly executed. Therefore,
The black region coordinate holding unit 108 stores the information of the black region obtained by the black region extraction unit 107 and the information of the black region obtained by dividing the black region by the black region division unit 110 as necessary. Will be retained.

FIGS. 11 (a) to 11 (c), FIGS.
FIG. 14 specifically shows what processing is actually performed on the target binarized image by the execution of the processing shown in FIG. 10 in the black area dividing unit 110.

First, the image shown in FIG. 11A has “123” and “4” on a white background, similarly to the image shown in FIG.
Normal areas where black characters such as "56" and "798" exist
This is an image that exists in a state circumscribing a reverse area where white characters “ABCDEFG” and “Irohanihohet” are present on a black background. When such an image is processed by the black region extraction unit 107, a black region (region indicated by the structure Group) in which the region indicated by the thick line in FIG.

Note that FIG. 11A shows the structure Part
FIG. 14 is an image diagram for showing a state in which s exists, and in fact, there are more structures Par than the number indicated by a thick line in FIG.
ts will be present.

Therefore, when the processing of step S1103 in FIG. 10 is executed in the black area dividing section 110,
Histogram histogram and histogram en
A histogram as shown in FIGS. 11B and 11C is obtained as dhist.

When the processing of the next step S1104 is executed, the histogram “sta” shown in FIG.
rhist, a dense point 1501 is detected, and a histogram endhis shown in FIG.
Congestion points 1502 and 1503 are detected from t.

In the next step S1105, a decision (first decision) is made as to whether or not to divide using the dense points in step S1104. Here, the dense points 150 of the histogram endhist are determined.
Since 2,1503 are in a distant relationship, it is determined that division is necessary. That is, the dense points 1502, 1
The fact that 503 is in a distant relationship means that the structure Par
This means that the density of “endx” indicating the coordinates of the pixel at the end of the set of black pixels indicated by ts is far away, so it is necessary to divide the black area indicated by the structure Group Is determined.

When the processing in the next step S1106 is executed based on the determination result in step S1105, "sta
“0” is set in “rtpoint”, and “endpoi” is set.
A dense point 1502 is set at “nt”.

In the next step S1107, the spread in the y direction is detected. Here, in the three areas indicated by 0 to start point start point to end point end point to the end, the area and the area indicate the same area.
ndpoint) and the area (endpoint to the end) in the y-direction are represented by the structure Parts
(Information indicated by “701a” in FIG. 7)
, Etc. to detect.

In the next step S1108, based on the result of detection of the spread in the y direction in step S1107,
Again, a determination as to whether or not to divide (a second determination) is made. Here, the region (0 to 0)
The “endpoint” is “15” shown in FIG.
04 "to" 1506 ", and the region (en
dpoint to the end) is “1504” shown in FIG.
Since it can be recognized that the data is distributed only to “1505”, it is determined that the data should be divided.

When the processing in the next step S1109 is executed based on the determination result in step S1108, the structure G
The black area indicated by the loop is the area 2
Divided into two areas. That is, as shown in FIG. 12, an inverted area (black area) in which white characters of “ABCDEFG” and “Irohanihoheto” are present on a black background is “AB
The area is divided into two areas, an area of (CDDEFG) and an area of (Irohanihohet).

In the area division in step S1109, the division shown in FIG. 12 may need to be performed as shown in FIG. 13 depending on the situation. In this case, for example, with reference to the color information of the original multi-valued image, the division as shown in FIG.
It is also possible to select whether to perform the division as shown in FIG.

Therefore, by dividing the black area of the image shown in FIG. 11 as shown in FIG. 12 (or FIG. 13), for example, if the binarization processing is executed for OCR, Also, as before, "123",
An area where black characters such as "456" and "789" are present on a white background is inverted despite being a normal area (an area where black characters are present on a white background) (see FIG. 23 above).
As shown in FIG. 14, "ABCDEFG"
Only the black area where the white characters of “Irohanihohet” are present is inverted, and binarization that can satisfactorily OCR all the characters such as “ABCDEFG”, “Irohanihohet”, “123”, “456”, and “789” is performed. It becomes possible.

When the process of step S1110 is executed, a new structure Group indicating a new black region obtained by dividing the black region indicated by the structure Group as described above is added to the black structure. It is held in the area coordinate holding unit 108. In addition, the structure Group that has not been divided
Is held in the black area coordinate holding unit 108 as it is.

In step S206 shown in FIG. 2, the area analyzing unit 109 determines whether or not the target black area is a character area for all the black areas held in the black area coordinate holding unit 108. If it is a region, whether or not it is an inverted region, the average pixel value of the background (average background color), the average pixel value of the character (average character color), the optimal binarization threshold used for binarization And other analysis information.

For example, the coordinate information of the black area held in the black area coordinate holding unit 108 is the same as the black area 30 shown in FIG.
If the coordinate information is 1 to 310, the area analysis unit 109
First, the histogram calculation unit 103 acquires histograms corresponding to the black regions 301 to 310, respectively. Then, the region analyzing unit 109 determines that the black region 3
From the corresponding histograms of 01 to 310, the black area 30
For each of 1 to 310, it is determined whether the target black area is a character area, if it is a character area, whether it is an inverted area, and if it is an inverted area, whether it should be inverted, Further, an optimum threshold value for binarizing the target black area is calculated, and these results (analysis results) are stored in the structure in various information (see FIG. 7 described above) held in the black area coordinate holding unit 108. Threshold information “Threshold” suitable for binarizing the area indicated by the Group, the structure G
The flag information "Chara" indicating whether the area indicated by the group is a character area or a non-character area (eg, an image area).
Flg ", flag information indicating whether or not the region indicated by the structure Group is a character region when the region is an inverted character region (a black background and a white character portion).
"InvertFlg", a flag indicating whether or not substantially the same object exists on the binarized image created with the threshold value one immediately before the threshold value that created the binarized image in which the region indicated by the structure Group is extracted. Information "SameAsMot
her "and pointer information" next "indicating the address of the next structure Group.

Here, it is desirable to omit the information on the child's black area from the histograms corresponding to the black areas 301 to 310 obtained by the histogram calculation unit 103 when performing the above-described various analyses. For example, the black area 30
As shown in FIG. 15, the black region 303 includes a black region 307 (a black region of a child having the black region 303 as a parent). 3
A better analysis result can be obtained by not using information on 07 (information on overlapping pixels).
As a processing method for this, there is a method using a map image, a method using coordinates, and the like. Here, a method using a map image will be described as an example.

A method using a map image:
The histogram calculation unit 103 uses a map image when generating a histogram of a black area. For example, when creating a histogram of the black region 310 extracted from the binarization result (the binarized image shown in FIG. 3D) with the binarization threshold 403 close to “0”, the target image ( A map image of the multi-value image of the document shown in FIG. 3A is prepared, and all the pixels constituting the map image are set to “
Then, the pixel portion of the map image corresponding to the pixel portion accessed when creating the histogram of the black region 310 is set to “1”.
In the map image, the pixel part not accessed is “0”
Since the accessed pixel portion is “1”, a histogram may be created using information on pixels whose pixel values are not “1”. By creating a histogram in the histogram calculation unit 103 according to such a method, the area analysis unit 109 can determine whether the pixel information overlaps even in the black area having the relationship shown in FIG. Since an accurate histogram of the black area can be obtained, an optimal analysis process can be performed.

In the above-described method using the map image, the flag information "SameAsMother" is set to "T
The black region that is “RUE (the same object exists)” may be configured not to be subjected to the binarization process in order to reduce the time for performing duplicate binarization in the final binarization process. In this case, for the black area, the analysis processing in the area analysis unit 109 is not performed, or even if the analysis processing is performed, “1” is set to the pixel portion corresponding to the black area in the map image. Do not set.

A specific example of the analysis processing in the area analysis unit 109 is as follows. For example, the average value average of the histogram of the target black area obtained by the histogram calculation unit 103 and its skew skew are

[0124]

(Equation 1)

If the absolute value of the resulting skew skew is larger than the binarization threshold at the time of obtaining the binarized image in which the target black area exists, it is determined that the target black area is a character area. Structure G that recognizes and indicates the target black area
group information "CharaFlg" to "TRU"
E ”. In this case, if the value of the skew skew is“ positive ”, the target black area is recognized as an inverted character area (a black background and a white character area). The flag information “InvertFlg” of the structure Group indicating the black area is set to “TRUE”. In addition, as a threshold suitable for binarizing the target black region in this case, a threshold value at which the background portion of the target black region is erased is obtained from the histogram of the target black region, and the result indicates the target black region. Structure Group information "T
threshold ".

As described above, the area analysis unit 109
In step S206 shown in FIG. 2, analysis is performed on all the black areas held in the black area coordinate holding unit 108, and the analysis result is written in the black area coordinate holding unit 108. The flag information of the structure Group "Inver
The following processing is further performed on the black area in which tFlg ”is“ TRUE ”.

First, the flag information “I” of the structure Group
A black area in which “nvertFlg” is “TRUE” (hereinafter, referred to as a “target black area”) is a character area and an inverted character area (a black background and a whitish character part). Therefore, in this processing, the multi-valued image (the image information held in the multi-valued memory 192) of the portion corresponding to the target black area is inverted, and the image after the inversion is shown in FIG. The processing of steps S202 to S206 is repeatedly executed (see the dotted arrow in FIG. 2). The inversion processing at this time may be, for example, to invert the multi-valued image itself or to invert the histogram obtained from the multi-valued image (the histogram obtained by the histogram calculation 103). Good, and the method is not limited.

More specifically, for example, in the black areas 301 to 310 (see FIGS. 3B to 3C) obtained from the multi-valued image shown in FIG.
6, 310 are portions corresponding to the inverted character area of the multi-valued image, and these black areas 301, 306, 310
Of the black regions 306 and 310, the structure Gr
The bin created with the threshold value one immediately before the threshold at which the binarized image (the binarized image shown in FIGS. 3C and 3D) from which the region (the black region) indicated by “up” is extracted is created. Flag information "SameAsMo" indicating whether or not substantially the same object exists on the binarized image (binary image shown in FIG. 3B).
Since “other” is set to “TRUE (exists)”, the present process is executed on the black region 301.

FIG. 16A shows an inverted multi-valued image area corresponding to the black area 301. This inverted image 3
For 50, first, when processing is executed by the histogram calculation unit 103 in step S202 of FIG.
A histogram as shown in FIG. 17 is obtained. In the next step S203, the threshold determination unit 104 causes the above-described FIG.
7, two binarization thresholds 451 and 452 are determined. In the next step S204, the binarization unit 105 obtains a binarized image as shown in FIGS. 16B and 16C. The binarized image shown in FIG.
An image obtained by binarization based on the binarization threshold 451, and the binarized image shown in FIG.
52 is an image obtained by binarization based on No. 52.

In the next step S205, the black areas 351 to 355 are extracted from the binarized image shown in FIG. Of these black regions 351 to 355, the black regions 351 to 354 have a quadrangular shape. Therefore, the region analysis unit 109 recognizes this, and the black regions 351 to 355 are recognized.
The flag information “ShapeRecta” indicating whether or not the shape of the black area portion of the structure Group indicating the “4” is square.
ngle "is set to" TRUE (square shape) ".
A black region extraction process from the binarized image shown in FIG. As a result, black regions 356 to 358 are extracted from the binarized image shown in FIG.

The black regions 356 to 358 correspond to the structure Gr.
The binarized image created with the threshold value one before the threshold value at which the binarized image (the binarized image shown in FIG. 16C) from which the region indicated by “up” (the black region) is extracted is generated. Since almost the same object exists on the (binarized image shown in FIG. 9B), the area analysis unit 109 recognizes this and sets the flag information of the structure Group indicating these black areas 356 to 358. SameAsMother "to" TRUE
(Exists) ".

Therefore, the flag information "SameAsM"
This processing is executed again on the black areas 351 to 355 in which “other” is not “TRUE (existing)”, and the area analysis unit 109 determines whether the area indicated by the structure Group is a character area or a non-character area. (CharaFlg) indicating whether the image is an image region or the like, and a structure Grou
When the area indicated by p is a character area, the flag information “InvertFlg” indicating whether or not the area is an inverted character area (an area with a black base and a white character part) and an area indicated by the structure Group The threshold is set to information “Threshold” or the like suitable for conversion into a value.

As described above, all the black regions are extracted, and the binarization threshold value for each black region, whether to invert the multi-valued image corresponding to the black region, and the like are determined.

It should be noted that flag information "ShapeRectangle" indicating whether or not the shape of the black area is a square is shown.
For example, when the target black area is an area (black portion) having a shape as shown in FIG.
(Square shape) ", and in the case of an area having a shape as shown in FIG. 4B," FALSE (not square shape) "is set.

In the above-described embodiment, the binary memory 106 is used to store the capacity of the binary image obtained by the binarizing unit 105 (the number of binary images corresponding to the number of binary thresholds obtained by the threshold determining unit 104). In such a configuration, it is possible to binarize a multi-valued image with a plurality of binarization thresholds and create a plurality of binarized images at once. This is effective in terms of processing speed, but requires a capacity that can hold the number of images equal to the number of binarization thresholds. However, with the following configuration, the binary memory 106 can have the capacity for one image and can save the memory.

First, the threshold value decision unit 104 decides the first binarization threshold value, and the binarization unit 105 executes binarization using the binarization threshold value, and the result (binary image) Is stored in the binary memory 106 having a capacity for one image. Next, the black region extraction unit 107 executes a black region extraction process on the binarized image in the binary memory 106. After that, the binarized image in the binary memory 106 is discarded. Next, the threshold determination unit 104 determines an initial binarization threshold. Binarization unit 1
05 denotes binarization at the binarization threshold value by the black region extraction unit 1
The process is executed only for the region corresponding to the black region extraction result in 07. Then, the black region extraction unit 107 executes a black region extraction process on the binarized image of the region corresponding to the black region extraction result.

Therefore, in the configuration described above, the capacity of the binary memory 106 only needs to have a capacity for one image.
It is possible to cope with a case where a large amount of binary memory cannot be possessed.

In the above-described embodiment, a histogram is created from all pieces of pixel information constituting a target multi-valued image, a plurality of binarization thresholds are determined from the histogram, and the binarization thresholds are determined. Is configured to acquire a plurality of binarized images from the image data. However, the present invention is not limited to this. For example, a histogram for a black region obtained by the black region A threshold may be determined to obtain a binarized image. With such a configuration, there is an advantage that division into smaller areas can be performed more favorably.

In the above-described embodiment, the binarized image is generated from the multi-valued image. However, the present invention is not limited to this. The present invention includes a configuration in which the image is generated and the area is finally divided from the image.

Further, an object of the present invention is to supply a storage medium storing program codes of software for realizing the functions of the host and the terminal of the above-described embodiment to a system or an apparatus, and to provide a computer (a computer) of the system or the apparatus. It is needless to say that the present invention can also be achieved by a CPU or an MPU) reading and executing a program code stored in a storage medium. In this case, the program code itself read from the storage medium implements the functions of the present embodiment, and the storage medium storing the program code constitutes the present invention. ROM, as a storage medium for supplying the program code,
Floppy disk, hard disk, optical disk, magneto-optical disk, CD-ROM, CD-R, magnetic tape,
A non-volatile memory card or the like can be used. By executing the program code read out by the computer, not only the functions of the present embodiment are realized, but also the OS and the like running on the computer perform actual processing based on the instructions of the program code. It goes without saying that a part or all of the above is performed, and the processing realizes the function of the present embodiment. Further, after the program code read from the storage medium is written to a memory provided in an extension function board inserted into the computer or a function extension unit connected to the computer, the function extension is performed based on the instruction of the program code. It goes without saying that a CPU or the like provided in the board or the function expansion unit performs part or all of the actual processing, and the processing realizes the functions of the present embodiment.

[0141]

As described above, in the present invention, when a multivalued image to be processed (such as an image obtained by reading a color document) is divided into regions, at least one Create a quantized image (such as a binarized image) and obtain an area including a pixel group of a predetermined value (a circumscribed rectangular area including a cluster of one-dimensionally continuous black pixels) from the quantized image. I do. Further, a histogram is obtained for the start pixel (start pixel) and the end pixel (end pixel) of the pixel group existing in the area, and the dense points of these start pixels and end pixels are detected. The start pixel and the end pixel existing in the region obtained by dividing based on the points are detected in other dimension directions, and based on the detection result, it is determined whether or not to actually perform the division, and the region is divided. It was configured to be. The attributes of all the regions obtained in this manner, for example, whether the target region is a character region, if the target region is a character region, whether the background is black and the character portion is a whitish inverted region, , The average pixel value (average background color), the average pixel value (average character color) of characters, and the optimal threshold value used when binarizing the target area are analyzed. Thereby, even if the multivalued image to be processed is obtained from a document in which various combinations of color parts such as a background and a character area part are present, optimal area division can be performed, and In particular, even if the image has an area that would include another unnecessary area if a circumscribed rectangle is taken, it is possible to divide the area satisfactorily and efficiently, and perform subsequent processing in units of rectangles. Can also.

[Brief description of the drawings]

FIG. 1 is a diagram illustrating a configuration of an image processing apparatus to which the present invention has been applied.

FIG. 2 is a flowchart for explaining the operation of the image processing apparatus.

FIG. 3 is a diagram for explaining an example of a multi-value image to be processed by the image processing apparatus and a binary image based on a plurality of binarization thresholds generated in a binarization process on the multi-value image. .

FIG. 4 is a diagram for explaining a histogram obtained from the multi-value image.

FIG. 5 is a flowchart illustrating a black region extraction process in the image processing apparatus.

FIG. 6 is a flowchart for explaining a process of detecting a set of black pixels in the black region extraction process.

FIG. 7 is a diagram illustrating an example of information stored in a black area coordinate storage unit of the image processing apparatus.

FIG. 8 is a diagram for specifically explaining the black region extraction processing.

FIG. 9 is a block diagram illustrating a configuration of a black area extraction unit that performs the black area extraction processing.

FIG. 10 is a flowchart illustrating a black area dividing process in the image processing apparatus.

FIG. 11 is a diagram for explaining an example of a binarized image to be subjected to the black region division processing, and a histogram of the first pixel and the last pixel of a set of black pixels obtained from the image; .

FIG. 12 is a diagram illustrating an example of a result of area division in the black area division processing.

FIG. 13 is a diagram for explaining another example of a result of area division in the black area division processing.

FIG. 14 is a diagram for describing a result of binarization for OCR in a rectangular unit based on a result of area division in the black area division processing.

FIG. 15 is a diagram for explaining a state in which black areas obtained by the black area extraction processing and the black area division processing are nested.

FIG. 16 is a diagram for explaining hierarchical black area extraction processing (black area extraction processing by inversion processing) in the black area extraction processing.

FIG. 17 is a diagram illustrating an example of a histogram in the above-described black area extraction processing (black area extraction processing by inversion processing).

FIG. 18 shows information stored in the black area coordinate storage unit.
It is a figure for explaining flag information "ShapeRectangle" which shows whether the shape of a black area part is a square.

FIG. 19 is a diagram for explaining a conventional image area dividing method.

FIG. 20 is a diagram for describing a luminance distribution state of a target image (a multi-valued image that is difficult to binarize) in the image area dividing method.

FIG. 21 is a diagram for explaining a conventional image area dividing method using an adaptive threshold binarization method.

FIG. 22 is a diagram for explaining an example of an image to be processed by the image area dividing method.

FIG. 23 is a diagram for describing a result of binarization for OCR in a rectangular unit based on a result of performing the image region dividing method on the image.

[Explanation of symbols]

 REFERENCE SIGNS LIST 100 Image processing device 101 Luminance conversion unit 102 Multi-valued memory 103 Histogram 104 Threshold value determination unit 105 Binarization unit 106 Binary memory 107 Black region extraction unit 108 Black region coordinate holding unit 109 Area analysis unit 110 Black region division unit

Claims (16)

[Claims] 1. An image processing apparatus for dividing an area of a multi-valued image, comprising: a quantizing means for acquiring a plurality of quantized images from the multi-valued image; and a quantized image obtained by the binarizing means. A region extracting means for extracting a region including a pixel group having a predetermined value, and a region obtained by the region extracting unit, based on a density of start pixels and end pixels of a pixel group present in the region. An image processing apparatus comprising: an area dividing unit that divides an image; and an area analyzing unit that analyzes an attribute of each area obtained by the area extracting unit and the area dividing unit. 2. The image processing apparatus according to claim 1, wherein said quantization means acquires a binarized image as said quantized image. 3. The image processing apparatus according to claim 1, wherein said quantization means acquires a quantized image of an area portion of said multi-valued image based on an area extraction result of said area extraction means. . 4. The image processing apparatus according to claim 1, wherein said quantization means acquires said plurality of quantized images based on a plurality of thresholds obtained from a histogram of said multi-valued image. . 5. An area extracting means, comprising: a detecting means for detecting, from the quantized image, a pixel group in which pixels having the predetermined value are one-dimensionally continuous; and coordinates of the pixel group detected by the detecting means. First holding means for holding information; second holding means for holding coordinate information of a pixel group already detected by the detection means; information in the first holding means; and the second holding means 2. The image processing apparatus according to claim 1, further comprising a combining unit configured to perform a combining process of the information in the image data. 6. The image processing apparatus according to claim 5, wherein said detecting means detects said area for every M lines on said quantized image. 7. The area extracted by the area extracting means includes a circumscribed rectangular area including a one-dimensionally continuous pixel group, and the area dividing means includes an area extracted by the area extracting means. A first detection unit that detects a dense point of the start pixel and the end pixel from a histogram of the start pixel and the end pixel of the pixel group included in the pixel group; and a division obtained based on a detection result of the first detection unit. For a start pixel and an end pixel of a pixel group present in the region, a second detection unit that detects a spread in a dimensional direction different from the one-dimensional direction, based on a detection result by the second detection unit, 2. A dividing means for executing region division.
An image processing apparatus as described in the above. 8. An image processing system in which a plurality of devices are communicably connected, wherein at least one of the plurality of devices is the image processing apparatus according to any one of claims 1 to 7. An image processing system having a function. 9. An image processing method for dividing a multi-valued image into regions, wherein at least one or more quantized images are created from the multi-valued image, and a cluster of black pixels is formed from the quantized image. And an area dividing step of dividing the black area obtained by the area extracting step on the basis of the density of the start pixels and end pixels of the cluster of black pixels existing in the black area. And an area analysis step of analyzing attributes of all areas obtained by the area extraction step and the area division step. 10. The image processing method according to claim 9, wherein said region extracting step includes a step of creating a binarized image as said quantized image. 11. The image processing method according to claim 9, wherein a part of the at least one quantized image is not a quantized image obtained from the entire surface of the multi-valued image. 12. The region extracting step includes: a black detecting step of detecting a one-dimensionally continuous black pixel cluster from the quantized image; and a coordinate information of the black pixel cluster detected by the black detecting step. A first holding step of holding; a second holding step of holding coordinate information of a cluster of black pixels already detected by the black detection step; a holding information by the first holding step; and the second holding The image processing method according to claim 9, further comprising a combining step of performing a combining process of the held information by the steps. 13. The image processing method according to claim 12, wherein the black detection step includes a step of executing the detection processing for each M lines on the quantized image. 14. The black area extracted in the area extraction step includes a circumscribed rectangular area including a cluster of one-dimensionally continuous black pixels, and the area division step is extracted in the area extraction step. A first detection step of detecting a dense point of the start pixel and the end pixel from a histogram of a start pixel and an end pixel of a cluster of black pixels included in the black region; and a detection result of the first detection step. A second detection step of dividing the black region based on the first region and detecting a spread of a start pixel and an end pixel of a pixel group existing in the divided region in a dimensional direction different from the one-dimensional direction; 10. The image processing method according to claim 9, further comprising a dividing step of dividing the black region based on a detection result in the detecting step. 15. A computer readable storage of a processing program for implementing the function of the image processing apparatus according to claim 1 or the function of the image processing system according to claim 8. A storage medium characterized by the above-mentioned. 16. A storage medium, wherein the processing steps of the image processing method according to claim 9 are readable by a computer.