JPH0127468B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a frame pattern identification method for recognizing a frame pattern in a document image.

Regarding recent facsimile machines, copy machines, etc.
Rather than just copying the original image itself, for example, it can extract and read the information in each specific area from multiple sheets of the original, and then combine the extracted image information to record the image. Systems with image editing functions such as these are being actively developed. In order to make use of such an image editing function, an important issue is how easily and accurately an area to be edited in a document image can be specified.

Conventionally, in order to specify the area of a document image,
A method has been adopted in which the original image is displayed once on the display surface and the area to be edited is specified using a light pen, etc., but this method makes the entire device complicated and large. It has the disadvantage of being stored away.

In addition, as a simple means for easily specifying a special area in a document image, two positions on the X-Y coordinates of the document placed on the contact glass on the document table are designated as addresses. A system has been developed in which a cursor mechanism is provided that can be used to specify an area surrounded by a set line on the X-axis and a line on the Y-axis as the area to be edited. There is. However, with such means, there is a large discrepancy in area designation, and it is difficult to simultaneously designate and judge a plurality of areas.

Furthermore, methods have been adopted in which the image information of a document read by a scanner is once stored in memory, and then the address of a specific area is specified and the information within that specific area is extracted. The disadvantage is that it requires a large amount of memory.

The present invention has been made in consideration of the above points, and
General office documents, such as ledgers and slips, often contain necessary information within a frame, and for example, it is necessary to create a customer list, goods list, etc. from several slips. Focusing on the fact that documents are often handled frame by frame, we have developed a method that is particularly useful when performing image editing such as specifying areas for each frame and extracting information within the frame. The present invention provides a frame pattern identification method that allows frame patterns to be recognized.

Hereinafter, one embodiment of the present invention will be described in detail with reference to the accompanying drawings.

The frame pattern identification method according to the present invention first scans and samples the document surface pixel by pixel using a scanner, and based on the binarized image information, first identifies the main characters forming the frame. Long line segments in the scanning and sub-scanning directions are extracted according to a certain logical judgment, and then a logical judgment is made as to whether or not the extracted long line segments constitute a frame.

(1) First, the long line segment extraction method will be explained. It is executed according to the algorithm described below.

a Logical sum Ai, Aj of multiple pixels (for example, about 5 pixels) in each direction of main scanning and sub-scanning based on the image information read and binarized by scanning and sampling the document surface pixel by pixel. (See FIG. 1; here, the main scanning direction is indicated by X, and the sub-scanning direction is indicated by Y.)

Ai=1-(1-Si,j)(1-Si,j _-1 )(1
−Si, j ₋₂ ) (1−Si, j ₋₃ ) (1−Si, j ₋₄ ) …(1) Aj=1−(1−Si, j) (1−S _i−1 , j)
(1-S _i-2 , j) (1-S _i-3 , j) (1-S _i-4 , j) ...(2) b Ai = 1 is continuous for n or more pixels in the sub-scanning Y direction. If so, that part is extracted as a long line segment candidate.

If Aj=1 continues for n or more pixels in the main scanning X direction, that portion is extracted as a long line segment candidate.

Note that setting n=70 here is sufficient for practical purposes.

c. Determine whether the extracted long line segment candidates are the same line as follows.

○B. Continuous long line segments exist on adjacent lines. ○B. Does the rear end of the segment of the previous line overlap the tip of the segment of the next line?
(See Figure 2.a), or the segment of the next line exists within m pixels from the rear end of the segment of the previous line (See Figure 2.b). This is sufficient for practical purposes.

d. Extract long line segments that satisfy the conditions of ○a and ○b above as constituent elements of the long line.

Figure 3a shows image information Sij that has been read and binarized by scanning and sampling the original surface pixel by pixel, and is converted into 4 bits in the sub-scanning Y direction according to the algorithm in a above. If the logical sum Aj is calculated for each pixel, the result will be as shown in FIG. Now, if each Aj = 1 line part in the main scanning Therefore, a long line segment J as shown in c in the figure is extracted.

(2) Next, it is determined whether the extracted long line segment satisfies the following conditions, and if the condition is satisfied, the long line segment is recognized as a component of the frame.

a. For a certain long line segment in the main scanning X direction, there are two long line segments in the sub scanning Y direction that intersect with it or have their tips within p pixels in the vicinity. b. Each of the two long lines in the sub scanning Y direction There must be one other long line segment in the main scanning X direction that intersects each pixel or exists within p pixels in the vicinity thereof. Note that setting p to approximately 15 is practically sufficient.

That is, as shown in FIG. 4, when long line segments J1, J2 in the main scanning X direction and long line segments I1, I2 in the sub scanning Y direction are extracted, the addresses of both ends of the long line segment J1 are _j1 , y _j1 ), (x′ _j1 , y _j1 ), and also J2
Let the addresses of (x _j2 , y _j2 ), (x′ _j2 , y _j2 ),
Also, the addresses at both ends of the long line segment I1 are (x _i1 ,
y _i1 ), (x _i1 , y′ _i1 ), and similarly, when the address of I2 is (x _i2 , y _i2 ), (x _i2 , y′ _i2 ), a y _i1 < y _j1 +p, y _i2 < y _j1 +p x _j1 -p<x _i1 <x' _j1 +p, x _j1 +p <x _i2 <x' _j1 +p and b x _j2 -p<x _i1 , x' _j2 +p>x _i2 y _j2 -p< When y′ _i1 , y _j2 −p<y′ _i2 , each long line segment J1, J2, I1, I
It is identified that the frame is constituted by 2.

Therefore, in the present invention, while scanning the document surface with a scanner and reading the image information, frame patterns are sequentially identified as described above. At this time, a number is assigned to each frame pattern in the order in which the frame patterns are identified as the image information is read, and later the number is specified to specify a specific frame area in the document. This makes it possible to easily perform image editing such as extracting image information within that specific frame.

Furthermore, FIG. 5 shows an example of a configuration for specifically implementing the frame pattern identification method according to the present invention described above, in which a long line segment candidate extracting section in the main scanning direction and a long line segment candidate in the sub scanning direction are shown. After determining whether each extracted long line segment candidate is on the same line according to the candidate extraction part and each extraction result of those extraction parts, it is determined whether a frame is formed by the long line segment. The frame pattern is recognized by making the following logical judgments as described above.
It is composed of CPU.

The long line segment candidate extraction section includes a binarizer 1 that binarizes the input image signal BS sent pixel by pixel from the scanner, and a binarizer 1 that binarizes the input image signal BS sent pixel by pixel from the scanner, and a binarizer 1 in each main scanning direction of the binarized image information. Line shift registers 21 to 24 that are continuously connected to each other for sequentially shifting lines, image information directly sent from the binarizer 1, and each line shift register 2
An OR circuit 3 that sequentially calculates the logical sum of five pixels in the sub-scanning direction according to the outputs of 1 to 24, and a counter 4 that counts the output "1" (corresponding to Ai = 1) of the OR circuit 3. When the output of the OR circuit 3 changes from "1" to "0", a latch circuit 5 latches the counted contents of the counter 4 up to that point, and a latch circuit 5 latches the counted contents of the counter 4 up to that point. ), and if the latch content is equal to or higher than the reference value, the digital comparator 6 determines that it is a long line segment candidate, and the latch content is determined according to the long line segment candidate determination output from the comparator 6. That is, a memory 7 that temporarily stores the length of a long line segment, and a long line segment position information generation circuit that generates position information of the long line segment in the main scanning direction at that time in accordance with the judgment output of the long line segment candidate output from the comparator 6. 8, and a memory 9 for temporarily storing the position information of the generated long line segment.

The long line segment candidate extraction section includes a 5-bit shift register 10 that sequentially shifts the output of the binarizer 1, and an OR circuit that sequentially calculates the logical sum of 5 pixels in the scanning direction according to each bit output. 11, a counter 12 having a memory function for counting the OR output "1" (corresponding to Aj = 1) and temporarily storing the counting contents until the OR output changes to "0"; A digital comparator 13 that compares the stored content with a preset reference value (corresponding to n) and determines it as a long line segment candidate if the stored content is greater than or equal to the reference value.
and a memory 14 for temporarily storing the contents of the counter 12, that is, the length of the long line segment, in response to the judgment output of the long line segment candidate output from the comparator 13; The long line segment position information generation circuit 15 generates the position information of the long line segment in the sub-scanning direction at that time, and the memory 16 temporarily stores the generated position information of the long line segment.

It goes without saying that if processing speed is not an issue, most of the operations can be executed on the CPU side, and a general microcomputer can be used as the CPU.

As described above, in the frame pattern identification method according to the present invention, a document image in which necessary information is written in a frame, such as a ledger or a slip, is read by scanning and sampling it pixel by pixel using a scanner. , and based on the binarized image information, take the logical sum of information for each plurality of pixels in the main scanning and sub-scanning directions, and if the line length of consecutive 1's is longer than a certain value, divide the line length into a long line segment. means for extracting as candidates; means for determining whether at least two adjacent extracted long line segment candidates in the main scanning or sub scanning direction are the same line, and determining a long line segment; The frame pattern is easily and easily identified by means of determining whether a frame is formed by each determined long line segment in the main scanning and sub-scanning directions. It has the excellent advantage of being able to reliably identify frame patterns in a document.

[Brief explanation of drawings]

Fig. 1 is a diagram showing information on five pixels each in the main scanning direction , 3rd
Figure a shows an example of binarized image information, Figure b shows extracted long line segment candidates, Figure c shows the long line segments, and Figure 4 shows four long line segments. FIG. 5 is a block diagram showing an example of a configuration for concretely implementing the frame pattern identification method according to the present invention. 1... Binarizer, 21-24... Line shift register, 4... Counter, 5... Latch, 6, 13... Digital comparator, 7, 9, 14, 16... Memory,
8, 15... Long line segment position information generation circuit, 10... Shift register, 12... Counter with memory function.

Claims (1)

[Claims] 1 Read the original image by scanning and sampling it pixel by pixel with a scanner, and based on the binarized image information, calculate the logical sum of information for each plurality of pixels in the main scanning and sub-scanning directions. , a means for extracting the line length as a long line segment candidate if the line length of consecutive 1s is longer than a certain value, and a means for extracting the line length as a long line segment candidate, and at least two adjacent in the main scanning or sub scanning direction of each extracted long line segment candidate are the same. means for determining whether a line is a line and determining a long line segment; and means for determining whether a frame is formed by each of the determined long line segments in the main scanning and sub-scanning directions. A frame pattern identification method.