JP3253201B2 - Image processing device and image type determination method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image processing apparatus for discriminating what kind of images can be classified into a plurality of document images used in a general office.
And an image type determination method .

[0002]

2. Description of the Related Art Generally, in registering a document in an electronic filing apparatus, a file structure is hierarchically constructed in order not only to register a document with a title but also to enable quick retrieval. In other words, the documents are classified and registered in boxes such as cabinets and binders using the same concept as a conventional paper document file.

When the number of documents to be registered is small, it does not take much time to perform the document classification work manually, but the document classification work is quite complicated in registering a large number of documents.

Furthermore, in the registration of documents in a fixed format such as a form, the boxes into which those documents are classified are determined in advance, and it is inefficient to manually perform this document classification work.

[0005] However, a method for automatically performing this document classification work has not been established, and in the current electronic filing apparatus, for any document, the document classification work must rely on humans.

[0006]

An image processing apparatus, such as an electronic filing apparatus, which needs to identify and determine the type of a document is not equipped with a method for automatically performing a document classification operation. Relies on human hands. Therefore, registering documents is cumbersome and time-consuming,
It was inefficient work.

According to the present invention , the type of the target image is determined in advance.
Automatically determine from multiple types of registered images
Image processing device and image type determination method
The purpose is to do.

[0008]

An image processing apparatus according to the present invention.
Is a binarized image signal obtained by binarizing each pixel of the target image
Output means, and one screen from the binarization means
Based on the binary image signal for each run
From the coordinates in the main scanning direction, the coordinates in the sub-scanning direction, and the run length.
Run-length encoding means for outputting different run information,
The run length threshold from the run information from the run length encoding means
Run elimination means for eliminating runs shorter than the
Based on run information from which short runs have been removed by the removal means
Labeling the connected runs as one area,
Main scan at the start of the labeled run
Direction coordinates and sub-scanning direction coordinates as the starting point of the circumscribed rectangle
The largest of the above labeled end points of the run
The coordinates in the main scanning direction and the coordinates in the sub-scanning direction are
Calculated as a point, and the start and end points of this circumscribed rectangle
Labeling and circumscribed rectangle extracting means for obtaining the size of the shape;
Each of the circumscribed rectangles by the labeling / circumscribed rectangle extraction means
Depending on the starting point and the size of the circumscribed rectangle, the width is large and the height is
A small circumscribed rectangle is determined as a horizontal ruled line area, and the horizontal width is short and the vertical width
Is determined as a vertical ruled line area, and the determined rectangle is
Ruled line area determination means that outputs the starting point of the shape and its length and type
And the number of horizontal ruled lines, vertical ruled lines, and horizontal ruled lines for each type of image
The starting point and its length of vertical lines and the starting point and its length of vertical ruled lines
Registration means and the ruled line area determination means
The number of horizontal rules, the number of vertical rules, and the horizontal rule of the target image obtained from
Starting point and length of line and starting point and length of vertical ruled line,
One type of image registered by the registration means
Number of horizontal ruled lines, vertical ruled lines, starting point of horizontal ruled line and its length
By comparing the starting point of the vertical ruled line with its length,
For each starting point of each horizontal ruled line area of the target image,
For each type of image registered by the registration means
And compare the start point of each horizontal ruled line to find the closest one.
The difference between the distance and the length between the starting points of interest.
The average of the squares is calculated using the horizontal ruled line comparison value for each of the above multiple types of images.
And calculate the starting point of each vertical ruled line area of the target image.
For each, one registered by the above registration means
For each type of image, compare the starting point of each vertical ruled line
Are close to each other, and the distance between
The average of the sum of squares of the differences in
Out calculated as a comparison value of the vertical ruled lines, each said plurality of types of images
Calculate the average of the comparison value of the horizontal ruled line and the vertical ruled line
The registered image for the smallest value of the calculated average value
Determining the type of the image as the type of the target image;
It is composed.

According to the image type determining method of the present invention, a target image
Output a binarized image signal obtained by binarizing each pixel of
Based on the binarized image signal for each screen
The coordinates in the main scanning direction and the sub-scanning
Outputs run information consisting of the coordinates of the direction and the run length.
Runs shorter than the run length threshold are removed from the run information
Runs that are linked based on run information from which
Is labeled as one region, and this labeled
Smallest main scanning direction coordinates and subruns at the start of a run
Find the coordinates of the inspection direction as the starting point of the circumscribed rectangle, and
Largest main scan direction end point of run
The coordinates of the mark and the sub-scanning direction are obtained as the end points of the circumscribed rectangle.
Calculate the size of the circumscribed rectangle from the start and end points of the circumscribed rectangle
The starting point of each circumscribed rectangle and the size of the circumscribed rectangle
A circumscribed rectangle with a large width and a small height is defined as a horizontal ruled line area.
Judgment is made and the circumscribed rectangle having a short width and a long height is defined as a vertical ruled line area.
The start point of the determined rectangle and its length and type are determined.
Horizontal ruled line of the target image obtained by this output
Number, vertical rule number, horizontal rule start point and its length and vertical rule
The starting point, its length, and one species registered in advance
Number of horizontal ruled lines, number of vertical ruled lines, starting point of horizontal ruled line
Compare the length of a rule and the start of a vertical rule with its length
Respectively, the starting point of each horizontal ruled line area of the target image
For one of the types registered by the registration means
Compare the start point of each horizontal ruled line for each image
Find the thing and the distance and length between each of its starting points
The squared mean of the difference between
Of the vertical ruled line area of the target image.
One type registered above for each of the starting points
The closest distance is compared with the starting point of each vertical ruled line for each image
Find the thing and the distance and length between each of its starting points
The squared mean of the difference between
Calculated as the comparison value of
The average value of the comparison value of the ruled line and the comparison value of the vertical ruled line is calculated.
The type of registered image for the smallest value of the calculated average value
Is determined as the type of the target image.

[0010]

According to the image processing apparatus of the present invention, each pixel of the target image is
Output a binarized image signal obtained by binarizing the
Run open based on the binarized image signal for each screen
Coordinates in the main scanning direction and coordinates in the sub-scanning direction as starting points
Outputs run information consisting of the target and run length.
Remove runs that are shorter than the run length threshold.
Based on the removed run information, one
Labeled as an area of this labeled run
The smallest coordinate in the main scanning direction and the smallest
The coordinates are determined as the starting point of the circumscribed rectangle.
Largest main scan direction coordinate and sub run at the end of a run
Find the coordinates of the inspection direction as the end point of the circumscribed rectangle.
The size of the circumscribed rectangle is calculated from the start and end points of the shape.
The width is large depending on the starting point of the circumscribed rectangle and the size of the circumscribed rectangle.
A circumscribed rectangle with a small vertical width is determined as the horizontal ruled line area, and the horizontal width is determined.
A circumscribed rectangle with a short and long vertical width is determined as a vertical ruled line area.
Outputs the starting point of the determined rectangle, its length and type, and
The number of horizontal ruled lines, vertical ruled lines,
Starting point and length of horizontal ruled line and starting point and length of vertical ruled line
And one type of image registered in advance
Number of horizontal ruled lines, number of vertical ruled lines, starting point of horizontal ruled line, its length and vertical
By comparing the starting point of the ruled line with its length,
For each starting point of each horizontal ruled line area of the elephant image,
For each type of image registered by the recording means
Compare with the starting point of the horizontal ruled line and find the one with the closest distance.
Squared square of the difference between the distances and lengths between the starting points subject to
Average as the comparison value of the horizontal ruled line for each of the above multiple types of images.
Calculate and start each of the vertical ruled line areas of the target image
For each of the registered image types
Compare the start point of the vertical ruled line to find the closest one,
Squared square of the difference between the distances and lengths between the starting points subject to
Average as the comparison value of the vertical ruled line for each of the above multiple types of images.
Calculated and compared with the horizontal ruled line comparison value for each of the above multiple types of images.
Calculate the average of the comparison values of the vertical ruled lines and calculate the calculated average
The type of the registered image for the smallest value of
The type is determined.

[0011]

[0012]

An embodiment of the present invention will be described below with reference to the drawings. FIG. 1 is a configuration diagram of an image processing apparatus for determining a document type according to the present invention. In FIG. 1, 1 is an input image signal, 2 is a binarizing means for binarizing an input signal,
21 is a binary image signal, 3 is run-length encoding means,
31 is a run length signal, 4 is a run removing unit, 41 is an image signal after run removal, 5 is a labeling / circumscribed rectangle extracting unit, 51 is a rectangular signal, 7 is a ruled line region determining unit, 61 is a ruled line region signal, and 7 is a ruled line region signal. Image determination means 8 is a determination signal.
Hereinafter, the document type determination method of the present invention will be described in detail.

An input image signal 1 read by an input device such as a scanner is converted by a binarizing means 2 into a binarizing threshold Th.
And a binarized image signal 21
(0 or 1) is output. The run-less encoding means 3 converts the image from the binarized image information 21 into run information.
The ruled line area information 61 is extracted based on the run information. First, the run removing means 4 removes short runs in the main scanning direction and the sub-scanning direction, respectively.

Next, the labeling / circumscribed rectangle extracting means 5 integrates (labels) the connected runs as one area based on the image signal output from the run removing means 4, and adds the run to each of the labeled areas. Find the bounding rectangle. The ruled line area determination means 6 outputs ruled line area information 61 based on the size of the rectangle. The image determining means 7 determines the type of image based on the ruled line area information 61.

Next, an embodiment of an image processing apparatus equipped with the document type determination processing of the present invention will be described. FIG. 1 is a schematic configuration diagram of an image processing apparatus according to this embodiment.
This image processing apparatus reads and inputs a general document document by a reading device such as an image scanner, inputs the data as, for example, digital data of 8 bits per pixel, determines the document type, and outputs it. Hereinafter, this embodiment will be described in detail.

In FIG. 1, reference numeral 1 denotes an input image signal. The input image signal 1 (f) is compared with a predetermined threshold value Th by the binarizing means 2 to perform a binarization process, and outputs a binarized image signal 21 (g). The binarizing means 2 includes a threshold memory for storing a threshold Th and a comparator for comparing the input image signal 1 (f) with the threshold Th. If the input image signal 1 is smaller than the binarization threshold Th, "0" is output as the binarized image signal 21 (g), and if larger, "1" is output. The comparison process is shown in equation (1).

G = 0: f <Th g = 1: f ≧ Th (1) FIG. 2 shows the configuration of the binarizing means 2. The binarizing means 2 is composed of an 8-bit comparator. 8-bit input pixel signal 1
Is compared with a preset 8-bit threshold value Th, and (1)
The binarized image signal 21 is output under the condition shown in the equation.

Next, run information is created from the binarized image signal 21 by the run-length encoding means 3. FIG. 3 is a diagram for explaining a run information extracting method, which will be described in detail below. FIG. 3A shows an original image. FIG. 3B is a bitmap representation of the original image, where each square corresponds to a pixel, the hatched portion is a black (1) pixel, and the rest is white (0).
Represents a pixel. In FIG. 3B, the first black pixel is located at the coordinates (2, 1) and is further connected rightward to the coordinates (4.1).

The pixel at the coordinates (1, 6) corresponds to the coordinates (5, 5).
6) is connected. In this way, the connection state of the pixels is replaced with information (run) in which the pixels connected in the right direction are grouped as shown in FIG. Further, these runs are output as information on the starting point and length of the run. For example, the first run has a start point (2, 1) and a length of 3, and the second run has a start point (1, 2) and a length of 1 and so on. As described above, the run length encoding unit 3 outputs the start point and the length of the run from the binary image signal 21 as run information.

FIG. 4 shows the configuration of the run-length encoding means 3. First, the first pixel is stored in the latch 300 at the first clock by the binary image signal 21. Subsequently, at the second clock, the AND gate 310 outputs “1” when the first pixel is white and the second pixel is black, and outputs “0” otherwise. That is, the output 311 of the AND gate 310
(Run start signal) is a signal indicating the start of the black run.

On the other hand, the AND gate 320 outputs "1" when the first pixel is black and the second pixel is white, and outputs "0" otherwise. That is, the output 321 of the AND gate 320
(Run end signal) is a signal indicating the end of the black run. The counter 330 is reset by the above-described run start signal 311 and is counted by the image clock 22 (CLK). The output signal 331 of the counter 330 is stored in the latch 360 by the above-mentioned run end signal 321. That is, the run length is stored in the latch 360.

Next, the counter 331 stores the image clock 2
2 (CLK) and vertical image clock 23 (VCLK)
Is input, and the current image position (X coordinate, Y coordinate) is counted. X coordinate 341 and Y coordinate 34 of run start
2 is stored in the latches 370 and 380, respectively, by the run start signal 311 described above. Selector 3
Reference numeral 90 denotes an X coordinate, a Y coordinate, and a run length of the run start stored in the latches 370, 380, and 360.
Select and output in the order of run length.
1 is output from the counter 350 at the timing of the run end signal 321.

Based on the run information calculated by the above procedure, the rule removing area information 4, the labeling / circumscribed rectangle extracting means 5, and the rule area determining means 6 output ruled area information 61. Hereinafter, a method of calculating the ruled line area information 61 will be described in detail.

First, the run removing unit 4 independently removes short runs or long runs from the run information obtained by the run length encoding unit 3. This procedure will be described in detail. FIG. 5A shows a run obtained by the run-length encoding means 3. First, in the removal of short runs, only those with a short run length are removed. FIG. 5B shows an example in which the run length 1 is removed, and the second to sixth runs are removed. In other words, this process removes runs that satisfy L <Ths (2), where L is the run length and Ths is the run length threshold of the short run.

FIG. 6A shows a document image more specifically, which is an example of a document composed of ruled lines and characters. When the short run removal processing is performed on FIG. 6A, the result becomes as shown in FIG. 6B. As can be seen from this figure, in this processing, vertical ruled lines and part of characters are removed. In this run removal processing, short runs are removed while long runs are removed in parallel. This processing removes runs that satisfy L <Th1 (3) when the run length threshold of a long run is set to Th1.

FIG. 7 shows an example of the long run removal processing. The image shown in FIG. 7B is output from the document image shown in FIG. 7A. FIG. 8 shows run removing means 4.
1 shows a circuit configuration.

First, the latches 410, 420, 430, 4
Run information 31 is sequentially input to 40, and a run length, a Y coordinate, an X coordinate, and a run length are stored every three clocks. The comparators 450 and 460 are, for example, 13 bits (A
3th edition, equivalent to 400 dpi (dot per inch) main scanning pixels), and the comparator 450 compares the threshold Ths and the run length based on the equation (2) and satisfies the equation (2). "0" is output as the determination signal 451 otherwise. That is, "1" is output only in the case of a run equal to or greater than the threshold Ths.

On the other hand, the comparator 460 compares Th1 with the run length based on the equation (3). If the equation (3) is satisfied, “0” is output, otherwise “1” is output as the determination signal 461. Is done. That is, "1" is output only in the case of a run equal to or smaller than the threshold Th1. The gate 470 is an OR gate, and the OR signal of the determination signals 451 and 461 is output as the determination signal 471.

480, 482, 484 are latches,
When the determination signal 471 is "1", the X coordinate, the Y coordinate, and the run length are stored, respectively. Counter 486 is latch 4
X, Y coordinates stored at 80, 482, 484,
A counter that generates a selection signal for sequentially selecting the run length. The selector 488 is latched 480, 482, 48 in the order of X coordinate, Y coordinate, and run length by the select signal 487.
4 is output as a run elimination signal 41.

Next, the labeling / circumscribed rectangle extracting means 5 integrates (labels) the connected runs as one area based on the image signal 41 output from the run removing means 4.
FIG. 9A shows a run obtained by the run length encoding means 3, and FIG. 9B shows the run information. FIG.
In (a), the first run is connected to the second run. Further, the first run and the third run are also connected. That is, the first, second, and third runs are all connected. In this way, the labeling / circumscribed rectangle extracting means 5 integrates (labels) the connected runs as one region. (C) of FIG. 9 is (a) of FIG.
Is output as a label (A) by labeling the image information shown in FIG.

The labeling / circumscribed rectangle extracting means 5 obtains the position and size of the rectangle circumscribing each of the labeled regions. From FIG. 9C, the label (A) is obtained by integrating the runs (1), (2), and (3). Labeling / circumscribed rectangle extraction means 5
Then, the starting points (x1, y1), (x2, y
2) Find the smallest x and smallest y of (x3, y3) as the starting point (x0, y0) of the circumscribed rectangle.
That is, the starting points of the n runs included in the target label are (x1, y1), (x2, y2)... (Xn, yn)
X0 = min (x1, x2,... Xn) (4) y0 = min (y1, y2,... Yn) (5) Also, the end point (x
e, ye) and calculate the size of the circumscribed rectangle (xs, ys)
Is calculated by the following equation.

Xs = xe−x0 + 1 (6) ys = ye−y0 + 1 (7) FIG. 11 shows circumscribed rectangle information 51 calculated by the above procedure for the run information shown in FIG.

FIG. 10 shows a labeling / circumscribed rectangle extracting means 5.
2 is an example of the circuit configuration of FIG. Output signal 4 of run removing means 4
For 1, the first line information (X coordinate, Y coordinate, run length) of the first line is stored in the memory 512 via the selector 510. The run information stored in the memory 512 is passed through the selector 514 to the comparator 520,
At 522, 524, 526, and 528, the information is compared with the information on the second line.

In the following description, the start position of the run on the first line is X0, Y0, the run length is L0, the start position of the run on the second line is X1, Y1, and the run length is L1.
First, the comparator 520 compares the start position X0 of the first run of the first line with the end position X1 + L1 of the first run of the second line.

At this time, if X0> X1, "1" is output as the comparison signal 521, and otherwise "0" is output.

Similarly, the comparator 522 outputs "1" as the comparison signal 523 if X0> X1 + L1, and "0" otherwise.

The comparator 524 outputs "1" as the comparison signal 525 if X0 + L0> X1 + L1, and "0" otherwise.

The comparator 526 outputs "1" as the comparison signal 527 if X0 + L0> X1, and "0" otherwise.

The comparator 528 outputs "1" as the comparison signal 529 if Y0> Y1, and outputs "0" otherwise.

On the other hand, the adder / subtracter 530 calculates X0 + L0-X1 + 1 or X1 + L1-X0 + 1 as an integrated run length when the runs are connected, and outputs X0, Y0, X
1, Y1, X0 + L0-X1 + 1, X1 + L1-X0 +
Outputs 1.

In the judgment table 540, the comparison signals 521,
Based on 523, 525, 527, and 529, six conditions are set, and selectors 514, 544 and memories 546, 548 are set.
And a run selection signal 543 are output. The memory selection signal 545 is a signal for selecting a memory 546 for internal operation (a memory for storing run information for one line) and a buffer memory 548 for outputting a labeling result, and is “0”.
In this case, the memory 546 is selected, and when "1", the memory 548 is selected.

The selection signal 542 is a 4-bit signal, and the lower 3 bits are determined under the following six conditions, and the upper 1 bit is the comparison signal 529 output as it is. First, the first condition is that the comparison signal 523 is “1”.
Under this condition, the run of the second line is not connected to the run of the first line, and the memory selection signal 541 is set to “0”.
As “000” as lower three bits of the selection signal 542.
Is output as the run selection signal 543. Second
Is a case where the comparison signal 521 is "1", the comparison signal 523 is "0", and the comparison signal 525 is "1". Under this condition, the run of the second line is connected to the run of the first line.
“0” as the memory selection signal 541 and the selection signal 542
"001" as the lower three bits of the run select signal 54
"3" is output as "1".

The third condition is that the comparison signal 521 is "1", the comparison signal 523 is "0", and the comparison signal 525 is "0". Under this condition, the run of the second line is the first line. Linked to the run, and “0” as the memory selection signal 541,
“010” is output as the lower three bits of the selection signal 542, and “0” is output as the run selection signal 543. The fourth condition is that the comparison signal 521 is “0”, the comparison signal 523 is “1”,
When the comparison signal 525 is “1”, the run of the second line is connected to the run of the first line under this condition, and “0” is set as the memory selection signal 541 and the lower 3
"011" is output as a bit and "1" is output as a run selection signal 543.

The fifth condition is that the comparison signal 521 is "0", the comparison signal 523 is "1", and the comparison signal 525 is "0". Under this condition, the run of the second line is the first line. Linked to the run, and “0” as the memory selection signal 541,
“100” is output as the lower three bits of the selection signal 542, and “0” is output as the run selection signal 543. The sixth condition is that the comparison signal 527 is "0".
This indicates that the run and connection of the line have been completed, and “1” is set as the memory selection signal 541 and the lower 3
"101" is output as a bit and "0" is output as a run selection signal 543. The selector 544 selects the selection signal 542
To output label information.

First, when the lower 3 bits of the selection signal 542 are "000", (X1, Y
1) Output (L1, Y−Y1 + 1 (where Y is the line number being processed)) as the label size. When the lower three bits of the selection signal 542 are “001”, Y0
When the smaller signal of Y1 and Y1 is Y2, the label start position is (X1, Y2) and the label size is (X1, Y2).
0 + L0-X1 + 1, Y-Y2 + 1). When the lower 3 bits of the selection signal 542 are “010”, (X1, Y2) is output as the label start position, and (L1, Y−Y2 + 1) is output as the label size. Selection signal 542
If the lower three bits of the label are “011”, the label start position is (X0, Y2), and the label size is (L0,
Y−Y2 + 1).

When the lower three bits of the selection signal 542 are "10
In the case of "0", the label start position is (X0, Y2),
The label size is (X1 + L1-X0 + 1, Y-Y
2 + 1) is output. When the lower three bits of the selection signal 542 are “101”, (X0, Y
0), and outputs (L0, Y−Y0 + 1) as the label size. The selector 544 is selected by the memory selection signal 541.
The label information output from the memory 546 or the memory 5
48 is selected.

Further, the run information 513 of the first line and the run information 41 of the second line are updated by the run selection signal 543. That is, when the run selection signal 543 is “0”, the run information 513 of the first line is updated to the next run. When the run selection signal 543 is “1”, the second
The run information 41 of the line is updated to the next run. The above processing is repeated until the end of one line.

When the processing of the first line is completed, the memory 5
The run information of the second line stored in 46 is stored in the selector 5
The above calculation is performed by the run information 513 of the second line and the run information 41 of the third line, which are stored in the memory 512 via the line 10. The above is repeated until the end of one page.

Next, the ruled line area judging means 6 outputs ruled line area information 61 from the circumscribed rectangle information 51. Generally, as a feature of the ruled line, the horizontal ruled line has a large horizontal width of the image and a small vertical width. Further, the vertical ruled line has a short horizontal width and a long vertical width of the image. Utilizing this, the width threshold Th1 and the height threshold Th
2 and set the circumscribing rectangle information 5 that satisfies the following conditions.
1 is extracted as a horizontal ruled line.

Xs> th1 and ys <th2 (8) Further, a horizontal width threshold Th3 and a vertical width threshold Th4 are set, and those satisfying the following conditions are extracted from the circumscribed rectangle information 51 as vertical ruled lines.

Xs <th3 and ys <th4 (9) An area that does not satisfy any of the above equations (8) and (9) is a non-ruled line area. FIG. 12 shows the flow of the following procedure.

FIGS. 6C and 7C respectively show the above-described run removing unit 4 and the labeling / circumscribed rectangle extracting unit 5 of the document images of FIGS. 6A and 7A. The horizontal ruled line area and the vertical ruled line area are extracted through the respective means of the ruled line area determining means 6.

FIG. 13 shows an example of the circuit configuration of the ruled line area determination means 6. First, the starting coordinates (x, y) and size (xs, ys) of the rectangle are input as the rectangle signal 51.
Latches 614, 612, 61 by non-illustrated clocks
X, y, and xs are respectively stored in 0. Comparator 62
At 0, 622, 624, 626, xs and Th respectively
1, ys and Th2, xs and Th3, and ys and Th4 are compared. When xs> Th1, the comparison signal 621 is "1".
Otherwise, the comparison signal 623 is “0” and when ys <Th2.
Is “1”, otherwise “0”, and when xs <Th3, the comparison signal 625 is “1”, otherwise “0”, ys
If> Th4, “1” is output as the comparison signal 627, otherwise “0” is output. In the judgment circuit 630, the comparison signal 621
When 623 and 623 are "1" or when the comparison signals 625 and 627 are "1", "1" is output as the judgment signal 631, and x, y, xs, and ys are stored in the latches 640, 642, 644, and 646, respectively. You.

When the comparison signals 621 and 623 are “1”, “0” is set to the comparison signal 6 as the ruled line type signal 632.
When 25 and 627 are "1", "1" is output.
The counter 650 has latches 640, 642, 644, 64
6 is a selection signal for sequentially selecting the information of No. 6 by the selector 660, and sequentially outputs x, y, xs, ys and types of rectangular signals determined to be ruled lines as the ruled line area signal 61.

Using the ruled line area signal 61 described above,
The image determining means 7 determines the type of the image. Hereinafter, the determination method will be described in detail. Using the ruled line area signal 61, determination is made from the position and length of the ruled line.

First, the number of horizontal ruled lines of the target image is nl, the starting point of the horizontal ruled line is (xl (i), yl (i)), and the length is 11 (i). Here, i is 1 to nl. The number of vertical ruled lines is np, and the starting point of the vertical ruled line is (xp
(J), yp (j)) and the length is lp (j).
Here, j is 1 to np.

Here, there are n types of images which have already been registered and which are to be determined. Among them, the number of horizontal ruled lines, the number of vertical ruled lines, the starting point and length of the horizontal ruled line, the starting point of the vertical ruled line and the length _{Nl m, Np m, (X1} m (k), Yl m
_{(K)), Ll m (} k), (Xp m (l), Yp m
(L)), and Lp _m (l) (provided that j is 1 to n, k
Is 1~Nl _m, l is 1~Np _m).

First, the starting points (xl (1), yl (2)) to (xl (nl), yl) of the nl horizontal ruled lines of the target image
(Nl)) for _{each, (Xl m (k),} Yl m
(K)), the one with the closest distance is obtained (k at this time is assumed to be k = o), and the square sum average Al _m of the difference between the distance and the length is obtained by the following equation.

[0059]

(Equation 1) Similarly Request Ap _m with respect to the longitudinal ruled line.

[0060]

(Equation 2) Furthermore, their average _Am

[0061]

(Equation 3) Is calculated as Determines that the A _m was calculated for all registered n kinds of images, the most the same type as the target image the image having a small value A _m. The image determination means 7 is a CP
U and a memory.

As described above, the image signal of the target pixel in the image to be processed is binarized by the binarizing means, and the black run length is calculated from the binarized image signal by the run-length encoding means. A short run is removed from the run length code by the run removing means, and based on the image signal from which the short run has been removed, a region of the connected runs is examined, and a rectangle circumscribing this region is labeled and a circumscribed rectangle is extracted. The ruled line area is determined by the ruled area determining means based on the size of the extracted rectangle, and the type of image is determined by the image determining means from the determined ruled line area.

Thus, the type of the target image can be determined from the information such as the position and length of the ruled line extracted from the target image.
Therefore, in an image processing apparatus such as an electronic filing apparatus that needs to identify and determine the type of a document, the document classification operation can be performed without relying on humans. As a result,
Document registration can be performed efficiently and efficiently. The present invention is not limited to the above embodiment, and various modifications can be made without departing from the spirit of the invention.

[0064]

As described in detail above, according to the present invention,
The type of the target image can be
Image processing device that can automatically determine from the image
It is possible to provide a position and an image type determination method . Therefore, in an image processing apparatus such as an electronic filing apparatus that needs to identify and determine the type of a document, the document classification operation can be performed without relying on humans. As a result, the document can be registered efficiently and efficiently.

[Brief description of the drawings]

FIG. 1 is an overall configuration block diagram showing an image processing apparatus according to an embodiment of the present invention.

FIG. 2 is a block diagram showing an example of a circuit configuration of a binarizing unit.

FIG. 3 is an explanatory diagram of a process of run-length encoding.

FIG. 4 is a block diagram showing an example of a circuit configuration of a run-length encoding unit.

FIG. 5 is an explanatory diagram of a run removing unit.

FIG. 6 is a diagram illustrating an example of ruled line area determination.

FIG. 7 is a diagram illustrating an example of ruled line area determination.

FIG. 8 is a block diagram illustrating an example of a circuit configuration of a run removing unit.

FIG. 9 is an explanatory diagram of a labeling unit.

FIG. 10 is a block diagram illustrating an example of a circuit configuration of a labeling unit.

FIG. 11 is an explanatory diagram of a circumscribed rectangle extracting unit.

FIG. 12 is an explanatory diagram of a ruled line area determination unit.

FIG. 13 is a block diagram illustrating an example of a circuit configuration of a ruled line area determination unit.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 ... Input image signal 2 ... Binarization means 21 ... Binary image signal 3 ... Run length encoding means 31 ... Run length signal 4 ... Run removal means 41 ... Image signal after run removal 5 ... Labeling and circumscribed rectangle extraction Means 51: Rectangular signal 6: Ruled line area determination means 61: Ruled line area signal 7: Image determination means 8: Determination signal

────────────────────────────────────────────────── ─── Continuation of the front page (51) Int.Cl. ⁷ Identification code FI G06T 7/60 200 G06T 7/60 200K Examiner Isao Matsuura (56) References JP-A-4-28685 (JP, A) JP-A-5-233873 (JP, A) JP-A-4-95178 (JP, A) JP-A-58-51372 (JP, A) (58) Fields investigated (Int. Cl. ⁷ , DB name) G06T 7 / 00-7/60 G06K 9/20

Claims (2)

(57) [Claims] 1. A binary image obtained by binarizing each pixel of a target image.
Binarization means for outputting a binarized image signal, and a binarized image signal for each screen from the binarization means.
Then, the coordinates in the main scanning direction as the starting point of the run start
A line that outputs run information consisting of coordinates in the scanning direction and run length
Run length encoding means and a run length from the run information from the run length encoding means.
Run removing means for removing runs shorter than the threshold, and run information from which short runs have been removed by the run removing means
And labeling run as one region based on
And the smallest runway at the start of this labeled run
The coordinates in the scanning direction and the coordinates in the sub-scanning direction are the starting points of the circumscribed rectangle.
The largest of the end points of the labeled run
The coordinates in the main scanning direction and the coordinates in the sub-scanning direction
Determined as the end point, circumscribed by the start and end points of this circumscribed rectangle
Labeling and circumscribed rectangle extraction means for finding the size of a rectangle
And the circumscribed rectangle by this labeling / circumscribed rectangle extraction means.
Depending on the starting point and the size of the circumscribed rectangle, the width is large and the height is
A small circumscribed rectangle is determined as a horizontal ruled line area, and the horizontal width is short and the vertical width
Is determined as a vertical ruled line area, and the determined rectangle is
Ruled line area determination means that outputs the starting point of the shape and its length and type
And the number of horizontal rules, the number of vertical rules, and the start
Register the point and its length, and the starting point of the vertical ruled line and its length.
Registration means, and the target image obtained by the determination of the ruled line area determination means.
The number of horizontal ruled lines, the number of vertical ruled lines, the starting point of the horizontal ruled line and its length
And the starting point of the vertical ruled line and its length, registered by the above registration means
Number of horizontal ruled lines and vertical ruled lines for each type of image
Number, starting point of horizontal ruled line and its length, starting point of vertical ruled line and its
By comparing with the length, for each of the start points of each horizontal ruled line area of the target image,
One type of image registered by the registration means
And find the closest one by comparing the starting point of each horizontal rule with
The difference between the distance and the length between the starting points of interest.
The average of the squares is calculated using the horizontal ruled line comparison value for each of the above multiple types of images.
, And for each of the starting points of each vertical ruled line area of the target image,
One type of image registered by the registration means
And find the closest one by comparing with the start point of each vertical ruled line.
The difference between the distance and the length between the starting points of interest.
The sum of squares is calculated using the vertical ruled line comparison value for each of the above multiple types of images.
And the horizontal ruled line comparison value and the vertical ruled line
Calculate the average value of the comparison values and calculate the minimum value of the calculated average value.
The type of the registered image for the value is determined as the type of the target image.
The image processing apparatus being characterized in that comprising a constant determining means. 2. A binary image obtained by binarizing each pixel of a target image.
And outputs a binarized image signal based on the binarized binarized image signal for each screen.
The coordinates in the main scanning direction and the sub-scanning
Run information including the coordinates of the direction and the run length is output, runs shorter than the run length threshold are removed from the run information, and the short runs are linked based on the removed run information.
Runs are labeled as one area , and the smallest main run at the start of this labeled run
The coordinates in the scanning direction and the coordinates in the sub-scanning direction are the starting points of the circumscribed rectangle.
The largest of the end points of the labeled run
The coordinates in the main scanning direction and the coordinates in the sub-scanning direction
Determined as the end point, circumscribed by the start and end points of this circumscribed rectangle
The size of the rectangle is determined, and the starting point of each circumscribed rectangle and the size of the circumscribed rectangle are used to determine the horizontal size.
A circumscribed rectangle with a large width and a small vertical width is determined as a horizontal ruled line area
And a circumscribed rectangle with a short width and a long length is determined as a vertical ruled line area.
And outputs the starting point of the determined rectangle and its length and type
The number of horizontal ruled lines and vertical ruled lines of the target image obtained from this output
Number of lines, starting point and length of horizontal ruled line, starting point and vertical ruled line
Length and one type of pre-registered image
Number of horizontal ruled lines, vertical ruled line, starting point of horizontal ruled line and its length
By comparing the starting point of the vertical ruled line with its length, the starting point of each horizontal ruled line area of the target image is determined.
Each horizontal ruled line for each type of registered image
Seek close and the most distance comparing the start, and the target of their
The average of the sum of squares of the difference between the distance and the length between the starting points
Calculated as a comparison value of the horizontal ruled line for each of the plurality of types of images, and for each of the starting points of the vertical ruled line regions of the target image,
One type of image registered by the registration means
And find the closest one by comparing with the start point of each vertical ruled line.
The difference between the distance and the length between the starting points of interest.
The sum of squares is calculated using the vertical ruled line comparison value for each of the above multiple types of images.
And the horizontal ruled line comparison value and the vertical ruled line
Calculate the average value of the comparison values and calculate the minimum value of the calculated average value.
The type of the registered image for the value is determined as the type of the target image.
Image type determination method comprising constant to it.