JPH0795811B2 - Facsimile image information processing method - Google Patents

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method of processing facsimile image information, and particularly to facsimile (FAX) communication and OCR.
The present invention relates to a processing method applied to an information input system or the like in which technologies are integrated to remove noise mixed in a communication line and extra image information included in a transmission document.

[0002]

2. Description of the Related Art Recently, an information input system that combines FAX communication and OCR technology has been developed, and is becoming widespread as a communication system for receiving and placing orders for products. In this communication system, the ordering party scans the ordering information entered by numbers or marks on the input sheet by FAX and transmits the image information to the ordering center side, while the ordering center side preliminarily stores the product type in a disk device or the like. Master information about customers, stock of goods, etc. is stored, and the order information is confirmed by analyzing the image information of the received input sheet.

With respect to the input sheet used in the above-mentioned system, an entry guide pattern as shown in FIG. 5 is formed in the number entry area of the sheet, and it is unrecognizable when the entered number is read by the OCR technology. There is a proposal to prevent misreading (utility model application filed on May 18, 1992 / title of device: mark sheet for character entry). That is, in the above-described entry guide pattern, the pattern 71 according to the 7-segment mode of the numeral is formed by the band line of the dropout color, and the small black reference points 72a to 72f are formed at the center of each knot of the pattern 71. Yes, each standard point 72a to 72f that can be clearly recognized by the person who entered
By connecting the lines with a line segment, it is possible to prevent arbitrary entry of bookkeeping numbers or individual numbers, and to make numbers that are easy to recognize by OCR technology.

Further, according to the ordering / ordering communication system using this input sheet, since the black reference points 72a to 72f are provided, it is possible to receive an order even if each orderer side does not stock the input sheet. A method can be adopted in which the input sheet original is transmitted by FAX from the center side and the order details are entered in the incoming original by numbers (hereinafter referred to as "turnaround method").

[0005]

By the way, general FA
Although it is a problem in X communication, noise is mixed in a line or terminal during image information transmission, and black pixel information of several pixels, which is not the original document image information, is scattered in the received image information. It is not uncommon to lose. This problem greatly affects character recognition, which is an important function of the communication system according to the ordering and ordering, as compared with the case of transmitting a normal document. That is, the input area for numbers and marks on the input sheet shown in FIG. 5 is at most about 5 mm × 7 mm, and if minute black pixel blocks due to noise are mixed in that area, it may cause misreading. Become. For example, FIG. 6 shows a state in which specific numbers are entered in the number entry area of the input sheet. In the figure, areas 73, 74, and 75 are black due to noise and dirt on the input sheet. When pixels are mixed, there is a possibility that "5" will be misread as "6", "4" as "9", and "1" as "7". In particular, when the above-mentioned turnaround method is adopted, since the orderer side uses the original document received by FAX as it is as an input sheet, the image information finally received by the ordering center side has two noises on the transmission line. It will be affected, and the probability of misreading will increase accordingly.

Further, the black reference points 72a to 72f on the input sheet have an effective function when numbers are entered, but they are unnecessary image information when character recognition is performed by the OCR technology. Depending on the method, it is necessary to perform character recognition while processing the reference points 72a to 72f as invalid information. That is, the stroke analysis method for checking whether or not there is a line element composed of black pixels between the adjacent reference points 72a to 72f has little effect, but the pattern matching method, the sonde method, and the outline of the character In other recognition methods such as the method using the feature (1), the reference points 72a to 72f become obstacles in the analysis, and it is necessary to remove the reference points and execute the recognition.

Further, when writing numbers and the like on the input sheet, pencils are generally often used, but at the time of writing, the pencil dust of the pencil scatters and adheres to other areas, which is about several pixels. It is often the case that it is read as a black pixel and causes misreading.

Therefore, according to the present invention, unnecessary minute image information generated due to noise mixed on a transmission line or dirt on a transmission document in the FAX communication is invalidated, image quality is improved, and image information received by FAX is improved. Was created for the purpose of reducing unrecognizable characters and misreading when performing character recognition by OCR technology.

[0009]

According to the present invention, in processing facsimile image information, an image expanded in the image information storage means is displayed.
Check each pixel while scanning information and detect black pixel
First procedure for starting processing in the case of doing, and the first procedure
Adjacent to the black pixel detected by
The second procedure of checking the pixels in a circle and the cycle of the second procedure.
If black pixels are detected in the second check, they are detected.
The black pixels are the orbits executed before the orbital check.
Third step of determining whether the black pixel has been detected by the check
And it is determined that it is not a black pixel detected in the third procedure.
Each time, the black pixel is set as a new reference black pixel.
Repeated the cycle check of the second procedure and the determination of the third procedure.
And a fourth procedure to be executed according to the fourth procedure.
The number of repetitions of the second procedure and the third procedure is N times (N is
When it becomes more than the integer set in the range of 2 or more)
The reference black pixel detected in the first procedure is regarded as effective image information.
None, if the number of repetitions is less than N, the above
The processing method of FAX image information is characterized in that the reference black pixel detected in the first procedure is regarded as invalid image information .

[0010]

In the present invention, the reference black pixel is detected in the first procedure.
After that, the surroundings of the reference black pixel are checked to make an adjacency.
Check if there are black pixels, and check if black pixels are present.
Then, the black pixel is set as a new reference black pixel in the second procedure and
The third procedure is executed, and the procedure of the second procedure is executed in the third procedure.
Detected by lap check executed before check
If it is determined that it is not the only black pixel, go to the fourth step.
Therefore, the second procedure and the third procedure are recursively executed. Therefore,
The number of repetitions N of the second procedure and the third procedure according to the fourth procedure is controlled.
If it does, the reference black pixel detected in the first procedure
And at least how many black pixels are chained
Wear.

Then, the number of repetitions N is set to the effective image information.
Corresponds to the size of the black pixel block (the number of black pixels) that can be regarded as
If this is done, the reference black pixel detected in the first step is the effective image.
Pixels to be processed as information or processed as invalid image information
It is possible to determine whether the pixel should be, the size of N
By selecting an appropriate value, the
It is possible to remove noise. Incidentally, in the present invention, “to be regarded as invalid image information” includes means for converting into white pixels.

[0012]

Embodiments of the present invention will be described in detail below with reference to FIGS. First, FIG. 2 shows a network configuration example of an ordering / ordering system in which FAX communication and OCR technology are integrated. In this configuration, each orderer to call orders center side from FAX1a~1n by itself has, in addition to the same aspect of the order contents as shown in FIG. 6 the product type and orderer code, etc. The completed image information of the input sheets 2a to 2n is transmitted to the order receiving center side via the telephone network or the FAX network 3. On the other hand, on the order center side, a FAX that accommodates multiple lines
The image information of the input sheets 2a to 2n transmitted by the adapter 4 is received, the encoded and compressed image information is decoded, the decoded image information is transferred to the image analysis device 5, and the segmentation and characters are transmitted. Perform recognition processing.
Then, the recognition result is notified to the computer 6, and the computer 8 collates the result with the master information stored in the disk 7 such as the product type and the customer, and if the conditions are satisfied, the order for each product is placed. The order confirmation in which the number of items is written is returned to the FAX 1a to 1n on the orderer side via the FAX adapter 4, and the order contents are printed by the printer.
Record output to 8. In addition, this system can adopt a turnaround method. That is, the ordering center side uses the scanner (or FAX) 9 to input the above-mentioned input sheets 2a-2.
The input sheet 10 similar to n is read and the manuscript image information is stored in the disk 7, and at the request of the orderer,
Alternatively, the original image information of the disk 7 is sent to the orderer's FAX 1a to 1n according to the time when the orderer places an order at the order center every day.
It is also possible that the orderer uses the received manuscript to place an order.

By the way, the image analysis apparatus 5 provided in the data processing system on the side of the order receiving center has a structure as shown in FIG. In the figure, 11 is an EPROM that stores a control program and an analysis program for character recognition, 12 is a RAM that stores image information storage areas and various management data, 13 is a DMA controller for communication, and 14 is a FAX adapter 4 Serial I / O for transmitting / receiving data to / from the computer, 15 a timer, 16 I / O receiving an interrupt signal from the computer 6, and 17 a computer 6
Side, an interrupt controller that manages interrupts from the serial I / O 14 and timer interrupts, 18 is a parallel I / O when outputting an interrupt request to the computer 6, and these modules are connected to the CPU 19 bus. C
The operation of the analyzer 5 is controlled by the PU 19 executing the program of the EPROM 11.

In this embodiment, the image analysis device is
The input sheet image information that the device 5 has decrypted from the FAX adapter 4
The feature is that the following pre-processing is executed when the information is received.
It The details of the pre-processing will be described below with reference to the flowchart of FIG.
And the image information matrix development diagram in Fig. 1
To do. However, here there are only 4 or less black pixels chained together.
Black pixel blocks that do not exist are regarded as invalid image information and black
A black pixel block in which five or more pixels are linked is used as effective image information.
A specific example will be given. In addition, in FIG. 1, hatching is applied.
Pixels that are marked as black pixels and pixels that are not shaded
Is expressed as a white pixel.

First, the FAX adapter 4 is
When the input sheet image information is received, it is decrypted and decrypted.
Image information is transferred to the real I / O14. Image analysis
In the device 5, the CPU 19 that has confirmed the transfer of the image information is DMA
The controller is activated and the image information is sequentially transferred to RAM12.
When the CPU 19 accumulates a predetermined number of lines, the CPU 19
Pre-processing program for PROM11 (part of analysis program)
To check the pixels of the accumulated image information in the scanning order.
Do it (S11, S12). That is, each line in FIG.
Check in the direction indicated by the arrow on the top row, and there are black pixels.
Check whether or not.

Then, black pixels are detected in the scanning process.
Then immediately, the black pixel count provided in the RAM 12 is counted.
Rewrite the flag (BP) from 0 to 1 and set the black pixel to the first
Adjacent pixels in the scanning direction as the reference black pixels for the eyes
The orbital check is performed clockwise from the pixel (S13 to S15). This
Here, black pixels are also detected by the above-mentioned circulation check.
If not, the black pixel detected in step S13 (first
(Reference black pixel of) is converted to white pixel, and BP rewritten to 1 is
Reset to 0 (S16 → S21, S22).

On the other hand, a black pixel is detected by the above-mentioned circulation check.
If this happens, rewrite BP to 2 and this time
The same black pixel as the second reference black pixel.
Check once (S16 to S20). Incidentally, the flow chart of FIG.
In the chart, the condition of step S17 is intervening.
The second reference black pixel in the step is the detected black pixel
Not applicable, so automatically around the second reference black pixel
Will be checked around.

Next, the periphery of the second reference black pixel
When a black pixel is detected as a result of checking the
, The detected black pixel is not the first black pixel.
Or not (S20 → S16, S17). That is, in this case
The first reference black pixel that has already been detected is
Since it is an adjacent pixel to the reference black pixel,
It will be detected in the check process, but the lap at this stage
The check is made before the first reference black pixel is detected.
Whether a black pixel adjacent to the second reference black pixel is detected
Will be checked. And the black image of the above condition
When a prime is detected, BP is incremented by 1
And set the detected black pixel to the third reference black.
The orbit check is executed as a pixel (S16 to S20).

Thereafter , similarly, the newly set standard is set.
When a black pixel is detected by the black pixel circulation check
, If the black pixel has not been detected, B
P is incremented by +1 and the black pixel is changed to a new base.
Repeat the procedure for executing the round-trip check as a quasi-black pixel (S
20 → S16 to S19 → S20). Therefore, the detected black pixel precedes
If it is not a black pixel detected by the procedure described above, BP is not +1
However, in this embodiment, BP is increased to 5
When it becomes, BP is reset to 0 and the first standard black
Complete the check sequence related to the pixel, then scan again
Check the pixels to see if black pixels are detected (S
19 → S22, S23 → S12). That is, BP is counted up to 5
That is, there are 5 or more blacks including the first reference black pixel.
The fact that the pixels exist in a chain has been confirmed.
Therefore, the first reference black pixel is regarded as one of the effective image information.
Then, leave the standard black pixel as it is and perform the next check
Move to Kens.

On the other hand, starting from the first reference black pixel
In the check sequence, the lap check before the BP becomes 5
Already detected by the procedure in which the black pixel detected by the
If it corresponds to, the first reference black pixel at that time
After converting to white pixels and resetting BP to 0,
Move to the next check sequence (S17 → S21 → S22, S23
→ S12). That is, in that case, the first reference black pixel is included.
The fact that there are less than four black pixels that are continuous in a chain
It means that the first standard black pixel is invalid.
It is regarded as one of the reports and converted into white pixels.

The check sequence is RA
For all pixels of page image information stored in M12
Although it is performed (S23 → S12), the process according to that sequence is illustrated.
When contrasted with 1, it becomes as follows. In the figure,
Black pixels indicated by " ○ " are white images in the above check sequence.
A black image indicated by " ● " to indicate that it is a black pixel that is converted to raw
The element is a black pixel that is left as it is.
First, the black pixels 51 and 52 do not have adjacent black pixels
In the same way as in the previous embodiment, the steps S16 → S21
Converted to plain. In addition, in this embodiment, black pixels 53 and 54
Comprising blocks, black pixels 55 to 58 or Ranaru blocks from
, 59 to 61 blocks composed of black pixels are also changed to white pixels.
This is different from the above embodiment in this respect. That is, Bro
Check the lap when black pixels 55 are detected.
, The black pixel 54 is immediately detected.
In the orbit check, except for the black pixel 55 detected earlier
Repeat procedure with BP = 2 because there are no black pixels
Is completed, and the block
Check the black pixels 56 → 57 → 58 in sequence
Will be executed, but based on the black pixel 58
Black pixels 55, 56, which were previously detected in the orbital check stage
Repeated with BP = 4 because there are no black pixels other than 57.
The return procedure is completed, and black pixel 59 is detected in the block.
After checking the orbit by, the order is based on black pixels 60 → 61
The next round check will be executed, but the black pixel 61
The black image that was detected earlier in the reference check stage
Since there are no black pixels other than element 60, BP = 3
This is because the iterative procedure ends. Also, this repeated hand
The maximum value of BP that accompanies in turn is which black image in each block
The same is true even if the lap check is started from the element.

Therefore, noise on the line or dust on the pencil lead
Therefore, 1 to 4 black pixel blocks are formed in the image information.
Even if you convert them to white pixels and unintentionally unnecessary
Image information can be removed and the image analysis device 5
Unnecessary in the analysis process related to character recognition
It is possible to prevent unrecognizable or misreading due to image information. Special
If the line is in poor condition or the above turnaround method
It is effective when adopting the formula. Also, according to this embodiment
If so, each guideline for the guide pattern shown in FIGS. 5 and 6
When points 72a to 72f correspond to 1 to 4 black pixel blocks
To perform character recognition as image information after removing it.
And become possible.

[0023]

The facsimile image information processing method of the present invention has the following effects due to the above configuration. In FAX communication, it mixes on the transmission line
Image information may have a few pixels due to
The black pixel blocks of some extent may be scattered, but the present invention
Is image quality by disabling a predetermined number or less of the black pixel blocks
And improve OCR from image information received by FAX.
When recognizing characters by technology, unrecognizable characters or misreading
Prevent life. Also, the number of repetitions of the second and third procedures
Large black pixel block to be invalidated by selecting N
The size can be adjusted, and it depends on the transmission conditions and character recognition conditions.
It has the advantage of being able to set appropriate noise removal conditions.
It For example, fax image information has a coarse linear density (fine mode).
In the normal mode, the sub-scanning line density is 7.7 lines / mm
In case of transmission / reception at 3.85 lines / mm)
Corrects image information distortion by selecting N according to the degree
Can be made, and the reference points printed with black dots are connected by line segments.
The image information of the input sheet that prevents the writing of arbitrary characters
When analyzing the information, remove the reference point in advance to obtain accurate characters.
It becomes possible to make recognition.

[Brief description of drawings]

FIG. 1 is a development view of an image information matrix.

FIG. 2 is a diagram showing a network configuration example of an ordering / ordering system in which FAX communication and OCR technology are integrated.

FIG. 3 is a system circuit diagram of the image analysis apparatus.

FIG. 4 is a flowchart showing a preprocessing procedure of image information.

[Fig. 5] A pattern for entering information in the number entry area on the input sheet
FIG.

[Figure 6] Enter specific numbers in the number entry area on the input sheet .
It is a figure showing the state where it is inserted.

[Explanation of symbols]

 1a to 1n FAX 2a to 2n Input Sheet 3 Telephone Network or FAX Network 4 FAX Adapter 5 Image Analysis Device 6 Computer 7 Disk 8 Printer 9 Scanner or FAX 10 Input Sheet 11 EPROM 12 RAM 13 DMA Controller 14 Serial I / O 15 Timer 16 I / O 17 Interrupt controller 18 Parallel I / O 19 CPU 31 to 39, 51 to 61 Black pixel ○ Black pixel converted to white pixel in preprocessing ● Black pixel left as black pixel in preprocessing

Claims (1)

[Claims] 1. A process of the facsimile image information, image
Each pixel while scanning the image information developed in the information storage means
Check and start the process when black pixels are detected.
The first procedure and the black pixel detected in the first procedure as a reference
Check the pixels adjacent to it as a black pixel
Black pixels are detected by the two steps and the round check of the second step.
If the detected black pixel is
The black image detected by the lap check executed before
3rd procedure to determine whether it is unqualified and detected in the 3rd procedure
Each time it is determined that it is not the only black pixel,
The round check of the second procedure using a new reference black pixel and
A fourth procedure for repeatedly executing the determination of the third procedure,
The second procedure and the third procedure according to the fourth procedure.
The number of times the sequence is repeated N times (N was set in the range of 2 or more)
If it is more than (integer), it was detected in the first procedure.
The reference black pixel is regarded as effective image information and repeated less than N times.
If it is the number of times, the standard detected in the first procedure
A method for processing facsimile image information, characterized in that black pixels are regarded as invalid image information .