JP3715674B2 - Image processing apparatus and ruled line recognition method - Google Patents

Description

[0001]
[Industrial application fields]
The present invention relates to an image processing apparatus for creating a document, and more particularly to an image processing apparatus having a means for extracting ruled lines included in an input image and a ruled line recognition method.
[0002]
[Prior art]
Many documents created by a character processing device include ruled lines such as tables.
[0003]
Conventionally, as a method for extracting a ruled line from an input image, a histogram representing the frequency of occurrence of black dots in the horizontal direction and the vertical direction of the input image is created and the ruled line is extracted.
[0004]
FIG. 5 is a diagram for explaining ruled line recognition according to the prior art. In the figure, (A) is an input image, and (B) is a histogram showing the frequency of appearance of dots in the input image (hereinafter referred to as black dots as a monochrome image) in the horizontal direction. In FIG. 5B, from the distribution of black dots, a1 to a3 are areas expected to be character parts, and b1 is an area expected to be ruled lines. The distinction between a character and a ruled line is based on the fact that the character part is wider in the vertical direction (vertical direction) than the ruled line, and the ruled line has a higher density (high frequency) of black dots than the character part.
[0005]
[Problems to be solved by the invention]
However, in the above conventional example, characters and ruled lines are determined based on only the frequency of black dots, and thus have the following drawbacks.
[0006]
(1) A part with a narrow vertical width and a high frequency is likely to be a ruled line, such as the part b1 in FIG. 5B, but if the character and the ruled line exist in the same line, the part a1 Ruled lines are not recognized due to wide width.
[0008]
Therefore, purpose of the present invention is to solve the drawbacks (1) described above, it is to provide an image processing apparatus capable of identifying and extracting the ruled lines mixed with characters.
[0009]
[Means for Solving the Problems]
In order to achieve the above object, an image processing apparatus according to the present invention includes a creating unit that creates a frequency distribution in the other direction at each position along at least one of a horizontal direction and a vertical direction with respect to an image. , At each position along at least one of the horizontal direction and the vertical direction, based on the slope calculation means for obtaining the slope of the frequency distribution and the slope value at each position obtained by the slope calculation means, In order from the position with the largest slope value, two positions having the same slope and opposite slope values are extracted as pairs, and the interval between the extracted pairs is narrower than the threshold value. And ruled line discriminating means for discriminating ruled lines in the image based on the pair .
[0011]
The ruled line determination unit creates a histogram for spacing of the pair, characterized in that there use a value between two peaks was in the histogram as the said threshold. In addition, the slope of the frequency distribution is obtained by differentiating the frequency distribution at each position.
[0012]
The image is bitmap data, the frequency distribution is created by counting each dot of the bitmap data, and the slope of the frequency distribution is obtained based on a change in the number of dots in the frequency distribution. And
[0013]
A ruled line recognition method according to the present invention is a ruled line recognition method in an image processing apparatus, and creates a frequency distribution in the other direction at each position along at least one of a horizontal direction and a vertical direction with respect to an image. A slope calculating step for obtaining a slope of the frequency distribution at each position along at least one of the horizontal direction and the vertical direction, and a slope value at each position obtained in the slope calculating step. Based on the position where the slope value is the maximum, the two positions having the slope values having the same absolute value and the opposite signs are extracted as pairs, and the interval between the extracted pairs is a threshold. And a ruled line discriminating step for discriminating a ruled line in the image based on a pair narrower than the value .
[0014]
[Action]
The present invention has means for differentiating the distribution of dots and obtaining the inclination thereof.
[0015]
When the distribution of the dots is differentiated and the inclination is obtained, the ruled line portion has a substantially symmetrical inclination at the beginning and end of the ruled line. This is because the ruled lines are basically composed of straight lines. On the other hand, the character portion has few characters that are symmetrical vertically and horizontally, and the dot distribution and its inclination are also asymmetric.
[0016]
The slope of the dot distribution in the ruled line portion of the same line as the character is steeper than that of the character, and a symmetric value appears at a position separated by the line width. Therefore, it is possible to extract ruled lines by paying attention to this point.
[0017]
【Example】
FIG. 1 is a block diagram showing the configuration of this embodiment. In the figure, a CPU is a central processing unit in the form of a microprocessor, which performs arithmetic and logical judgments for image processing, and is connected to these buses via an address bus AB, a control bus CB, and a data bus DB. Control each component. The CPU also creates a histogram from input image data, creates data obtained by differentiating the histogram, extracts ruled lines / character strings, discriminates ruled lines / character strings, and the like based on the contents of a RAM, which will be described later. The address bus AB transfers an address signal indicating a component to be controlled by the microprocessor CPU. The control bus CB transfers and supplies control signals for each component of the microprocessor CPU. The data bus DB transfers data between the component devices.
[0018]
Next, the ROM is a read-only fixed memory, and stores a control procedure by a microprocessor CPU described later.
[0019]
The RAM is a writable random access memory, and is used for temporary storage of various data from each component. Data to be stored include histogram data of black dots in both horizontal and vertical directions, slope data obtained by differentiating the histogram data, data indicating the position of the extracted ruled line / character string, and threshold data for distinguishing the ruled line from the character string and so on.
[0020]
SCAN is an image scanner, and IBUF is an input image buffer memory. Image information input from the image scanner SCAN is stored in the IBUF.
[0021]
CR is a cursor register whose contents can be read and written by the CPU. A CRT controller CRTC, which will be described later, displays a cursor at a position on the display device CRT corresponding to the address stored here.
[0022]
The CRT is a display such as a cathode ray tube and displays the contents stored in the input image buffer memory IBUF using a CRT controller CRTC.
[0023]
The CRT controller CRTC displays the contents stored in the cursor giresta CR and the buffer memory IBUF on the display device CRT.
[0024]
CG is a character generator that generates character signals such as characters, symbols, and cursors to be displayed on the display CRT.
[0025]
The image processing apparatus according to the present invention composed of such components operates in response to various inputs from the keyboard KB. When an input from the KB is supplied, an interrupt signal is first sent to the CPU. The CPU issues various control signals stored in the ROM, and various controls are performed by the control signals.
[0026]
Next, FIG. 2 shows a flowchart of an example of ruled line extraction.
[0027]
In step S1, an image to be discriminated is input as bitmap data.
[0028]
In step S2, a histogram (frequency table) of black dots in the horizontal direction (lateral direction) of the bitmap data is created.
[0029]
In step S3, for each position in the vertical direction (vertical direction), a change in the frequency value in the horizontal direction is differentiated to create a table with the change in the number of black dots as the slope.
[0030]
In step S4, ruled line candidates are extracted from the slope table. In the case of a ruled line, since it is composed of straight lines, the slope is steep and the slope of the beginning of the ruled line and the slope of the end are almost the same. In the case of horizontal writing in FIG. 3, an image in which characters and ruled lines are present in the same line (FIG. 3A, its histogram (FIG. 3B)) and a value obtained by differentiating the frequency distribution of black dots in the horizontal direction (FIG. 3 ( 3C, c1 and c2 in FIG.3C are symmetrical values at the slopes of the beginning and end of the ruled line, and the interval between the pair of slopes is the thickness of the ruled line. The one with the largest value (steep slope) in the numerical table created in step S3 is the beginning of the ruled line, and the part where the absolute value is almost the same and the sign is reversed appears after several dots after that. It is assumed that there is a value of this pair as one ruled line candidate.
[0031]
In step S5, step S4 is repeated until the extraction in step S4 can no longer be performed (the start and end positions are close to the vertical width of the read image).
[0032]
In step S6, a threshold value for determining the ruled line and the character string is obtained from the extracted value of the ruled line candidate pair. A specific example of how to obtain the threshold value will be described with reference to FIG. As shown in FIG. 4 (A), of the paired value intervals (line widths) extracted as ruled line candidates, a narrowly spaced portion d1 is a ruled line, and a wide portion d2 is a character string. This inclination interval is the thickness of the ruled line candidate. Therefore, when a “thickness” histogram is created for each ruled line candidate, as shown in FIG. 4B, there are peaks around the height of the character (e2) and around the thickness of the ruled line (e1). A histogram is created. The value between these two peaks, usually the median value (e3) is taken as the threshold value. However, since the thickness of the ruled lines and the size of the characters differ depending on the character size of the read document, the threshold values for determining “narrow” and “wide” also differ depending on the document.
[0033]
FIGS. 4C and 4D are histograms of candidate thicknesses when the character size is small and large, respectively, where f1 and g1 are narrow pair candidates, and f2 and g2 are pair widths. It is a broad candidate histogram. The median values f3 and g3 are set as threshold values. Thus, a threshold value is obtained for each document.
[0034]
In step S7, among the ruled line candidates extracted in step S4, those that are narrower than the threshold obtained in step S6 are determined as the final ruled line candidates in the horizontal direction.
[0035]
In step S8, a histogram of black dots in the vertical direction of the bitmap data is created.
[0036]
In step S9, a gradient number table obtained by differentiating the histogram is created, ruled line candidates are extracted in step S10, repeated in step S11, threshold values are obtained in step S12, and final vertical ruled line candidates in step S13. To decide. Details are the same as those in the horizontal direction, and the description is omitted. The method for obtaining the threshold value in the vertical direction is the same as the method for obtaining the threshold value in the horizontal direction in step S6. However, the value obtained in step S6 can also be used.
[0037]
【The invention's effect】
As described above, according to the present invention, since the ruled line is formed of a straight line, the ruled line can be easily discriminated using the fact that the inclination is steeper than that of the character string. That is, a ruled line that tends to be mistaken for a character string can be extracted from the ruled line on the same line as the character string. In addition, the ruled line inclined by skewing tends to be lower in the height of the histogram than the character string, but can be extracted because the histogram has a steep inclination.
[Brief description of the drawings]
FIG. 1 is a configuration diagram of an embodiment of the present invention.
FIG. 2 is a flowchart of processing of an embodiment.
FIG. 3 is an explanatory diagram of ruled line extraction;
FIG. 4 is a diagram illustrating a specific example of how to obtain a threshold value for ruled line discrimination.
FIG. 5 is a conceptual diagram of ruled line recognition according to a conventional technique.
FIG. 6 is a diagram illustrating a problem of ruled line recognition according to a conventional technique.
[Explanation of symbols]
CPU Central processing unit SCAN Image scanner IBUF Input image buffer memory RAM Writable random access memory ROM Read only fixed memory

Claims (5)

Creating means for creating a frequency distribution in the other direction at each position along at least one of the horizontal direction and the vertical direction with respect to the image;
An inclination calculating means for obtaining an inclination of the frequency distribution at each position along at least one of the horizontal direction and the vertical direction;
Based on the value of the inclination at each position obtained by the inclination calculating means, two positions having an inclination value with substantially the same absolute value and opposite signs, in order from the position having the largest inclination value. And a ruled line discriminating means for discriminating ruled lines in the image on the basis of a pair whose extracted pair interval is narrower than a threshold value . The ruled line determination unit creates a histogram for spacing of the pair, the image processing according to the value between two peaks was in the histogram to claim 1, characterized in that there use as a said threshold apparatus.   The image processing apparatus according to claim 1, wherein the slope of the frequency distribution is obtained by differentiating the frequency distribution at each position. The image is bitmap data, the frequency distribution is created by counting each dot of the bitmap data, and the slope of the frequency distribution is obtained based on a change in the number of dots in the frequency distribution. The image processing apparatus according to claim 1 . A creation step of creating a frequency distribution in the other direction at each position along at least one of the horizontal direction and the vertical direction with respect to the image;
An inclination calculating step for obtaining an inclination of the frequency distribution at each position along at least one of the horizontal direction and the vertical direction;
Based on the slope value at each position obtained in the slope calculating step, the two positions having the slope values having the same absolute value and the opposite signs in order from the position having the largest slope value. And a ruled line discrimination step for discriminating ruled lines in the image based on pairs in which the interval between the extracted pairs is narrower than a threshold value .

Images