JPS5917473B2 - Character area determination method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to an optical character reading device, and more particularly to a character area determination method that can accurately determine the area of each character.

A typical optical character reading device uses a reading head in which a large number of photoelectric conversion elements are arranged in a line to sufficiently cover the height of the characters, and this head is moved at a constant speed in the direction of character alignment (character line direction). The character image on the form is converted into a character pattern signal by photoelectric conversion while transmitting the image.

Further, the reading head is fixed, the form is wound and fixed around a rotating drum, and the rotating drum is rotated at a constant speed while photoelectric conversion is performed by the reading head to convert it into a five-character pattern signal.

Further, in a normal character reading device, it is necessary to determine the area of each character (separate each character).

For this purpose, in the past, the edge of the paper or a pre-printed marker was detected, and using this as a reference, a reference timing signal (hereinafter referred to as the normal reference timing signal) was sent in response to the relative feed of one character pitch. ), and each character area was separated at the position corresponding to this pulse signal.

Therefore, errors may occur in the character area due to expansion and contraction of the paper or written characters that are outside the character frame, which causes a reduction in the character recognition rate.

An object of the present invention is to reduce the error when determining a character area. In the present invention, in a character pattern signal, a group of bits in a direction orthogonal to a character line is treated as a column, and a white column identification signal is created by detecting a column in which the number of black bits in each column is equal to or less than a reference number as a white column.

Further, in the present invention, a search area signal indicating a search area sufficient to correct an error in a character area due to expansion and contraction of paper, etc. is set based on the normal reference timing signal.

Initially, it matches the positive reference timing signal,
Thereafter, a tentative reference timing signal is sequentially created to be updated as described below. Then, the white line identification signal is searched in each search area to detect the white line closest to the tentative reference timing signal, and the character area is separated at the position of this white line. Or search for the white row identification signal in two adjacent search areas,
Pairs of white columns are detected at positions separated by a character pitch, and character areas are separated at the position of the pair of white columns closest to the tentative reference timing signal. In order to separate the next character area, the next temporary reference timing is updated and set at a position that is one character pitch away from the position of one of the white rows or a pair of white rows.
This will be explained in detail below using the drawings. FIG. 1, FIG. 2, and FIG. 3 are explanatory diagrams of a first embodiment of the present invention.
is the edge of the paper, 2 is the character frame, 3A, 3B, 3Z are the pulses of the standard timing signal S3 at that time, 4-1A,
4-1B is a temporary reference timing signal S4- at the initial stage.
1 pulse, 4-2A, 4-2B, 4-3A, 4-3B
are updated temporary reference timing signals S4-2 and S4-3
, 5-1 and 5-2 are the detection areas at that time, 6-1 and 6-2 are the determined character areas, S1 is the sampling pulse signal, and S2 is the pulse that detected the edge of the paper. The signal S5 is a white row identification signal, and S6 is a signal indicating the search area.

First, referring to FIG. 1, a normal reference timing signal S3 and a first provisional reference timing signal S4-1 are set based on the signal S2 that detects the edge 1 of the paper (pulse 4-
1A to match pulse 3Z), and indicates a search area sufficient to correct errors in the character area due to expansion and contraction of paper, typefaces written outside the character frame, etc., centered around each pulse of the positive reference timing signal S3. Determine signal S6.

Now, for example, when deciding the frame of the character area for character [3],
Two rows separated by the pulse interval of the positive reference timing signal S3 are simultaneously looked at from one end of each of the two search areas 5-1 and 5-2, which are separated by the pulse interval of the positive reference timing signal S3, and children are determined one after another. The presence or absence of a row of black bits (hereinafter referred to as a white row) that is less than or equal to the specified reference number is detected in the search areas 5-1 and 5-2.
Explore to the other end. As a result, as shown in FIG. 1, in two rows separated from one end of each of the two search areas 51 and 5-2 by the pulse interval of the positive reference timing signal S3, two rows (hereinafter referred to as This is called a first area candidate sequence.

), it is determined that the characters are separated at that location, and the pulse 4-2A of the provisional reference timing signal S4-2 at that point in the first region candidate sequence is
The first region candidate sequence closest to 4-2B is set as the front end and rear end of one character region, and the character region 6-2 is determined at the timing of the sampling pulse signal S1. And the pulses 4-2A, 4-2 of the current temporary reference timing signal S4-2
If the position of the first area candidate sequence closest to B does not match the pulses 4-2A and 4-2B of the current provisional reference timing signal S4-2, the position of the first area candidate sequence is newly changed to the next runner-up. Pulse 4-3A, 4 of temporary reference timing signal S4-3
-3B.

Pulses 4-2A, 4- of temporary reference timing signal S4-2
When 2B is updated, a gap or overlap occurs between the previous character area 6-1 and the current character area 6-2.

Fig. 2 shows that as a result of searching the search areas 5-1 and 5-2 for character [3], there is no first area candidate sequence, and the search area 5
-1 and 5-2 are two columns with a white column only on one side (hereinafter referred to as the second area candidate column).

) exists, and as in the case where the first area candidate sequence in FIG. The second area candidate string is set as the front end and the rear end of one character area, respectively, and the character area 6-2 is determined at the timing of the sampling pulse signal S1. However, unlike the case where the first area candidate sequence in FIG. 1 exists, the provisional reference timing signal S4-2 at that point in time is not updated. FIG. 3 shows a case where, as a result of searching the search areas 5-1 and 5-2 of character [3], neither the first area candidate sequence nor the second area candidate 1 sequence exists.

In this case, the temporary reference timing signal S4 at that time
The two rows of pulses 4-2A and 4-2B of -2 are the front and rear ends of one character area, and the sampling pulse signal S
Character area 6-2 is determined at timing 1. Also, the second
As in the case where the second area candidate sequence in the figure exists, the provisional reference timing signal S4-2 at that point is not updated. Similarly, the next character [4] is also a temporary standard determined by the correct reference timing signal S3 that uniquely determines the character area from the edge 1 of the paper and the error due to paper expansion and contraction, etc. by determining the area of the previous character. Pulse 4-3A of timing signal S4-3,
The character area is determined by 4-3B.

In the above explanation of FIGS. 1 to 3, character separation areas are determined by searching two search areas based on the priorities of the first and second area candidate sequences, but It is also possible to determine the character separation area by

In other words, in the search area, if a white column exists, the white column closest to the pulse of the temporary reference timing signal is used to separate the characters, and if there is no white column, the pulse train of the temporary reference timing signal is used to separate the characters. By performing only separation and updating the provisional reference timing signal as described above when separation occurs at a white column, the character area can be easily determined. FIG. 4 is a block diagram of the first embodiment of the present invention, where 10 is a one-character buffer, 11 is a character buffer during search, 12-1 and 12-2 are blank column identification circuits, and 13 is a positive reference timing signal. 14 is a control circuit; 15 is a multiplexer for assisting with overlapping or gaps in character areas;
16 is a column counter indicating the column being searched, 17 is the first and second
A multiplexer for region candidate strings other than region candidate strings; 18 is a register for storing region candidate strings (hereinafter referred to as candidate string register); 19 is a register for storing determined character region strings (hereinafter referred to as determined string register). ), 20
2 is a multiplexer for setting an initial value, 21 is a register (hereinafter referred to as a temporary reference register) for storing a pulse train of a temporary reference timing signal, and 30 is a processing block.

In FIG. 4, the binarized input character data D1 is 1
One character is stored in the character buffer 10, the character buffer 11 at the time of the search is searched from one end of the search area determined based on the correct reference timing signal, and the determined string register 1 is stored.
Output data D2 is output from the multiplexer 15 in accordance with the contents D6 of 9.

Note that a signal indicating the exploration area (signal S6 in FIGS. 1 to 3) is generated in the control circuit 14 based on a positive reference timing signal. On the other hand, the signal S that detected the edge of the paper
2 operates the positive reference timing signal generator 13 at the timing of the sampling pulse signal S1 to generate the positive reference timing signal S3, and at the same time sets the initial data D3 for the temporary reference timing signal in the temporary reference register 21.

Normally, the initial data D3 is made to match the pulse train of the positive reference timing signal S3. Moreover, the white column identification circuit 12-1, 1
2-2 determines the black and white information of the columns before and after the one-character buffer 10, and sends white column identification signals S5-1 and S5-2 of the two search areas to the control circuit 14 at the timing of the sampling pulse signal S1.

The control circuit 14 outputs the normal reference timing signal S3, the white row identification signals S5-1 and S52, and the content S of the temporary reference register 21.
4, the next control signals SlO to Sl4 are outputted by the method explained in FIGS. 1 to 3. That is,
A column count signal SlO that counts each column from one end of the search area, and a valid data set signal Sl that becomes the area candidate column.
3. For setting the update signal Sll of the temporary reference register 21 when the area is determined by the first area candidate string, and the content S4 of the temporary reference register 21 when the area is other than the first and second area candidate strings. It outputs a switching signal Sl2 and a character start signal Sl4 when determining a character area.

Furthermore, the control circuit 14 and its control signals SlO to Sl4
To explain, first, the column counter 16 is operated by the column count signal SlO to count columns from one end of the search area, and at the same time, the white column identification signals S5-1 and S5 are activated.
-Explore the white row of 2.

If there are first and second region candidate columns as a result of the search, a valid data set signal Sl3 is output so that the region candidate columns become the region candidate columns closest to the content S4 of the temporary reference register 21, and the contents of the column counter 16 are D4 is set in candidate column register 18 through multiplexer 17. In this way, the search area is searched, and if the first and second area candidate strings do not exist in the search area at the end of the search, the content S4 of the temporary reference register 21 is transferred to the candidate string register 18 by the switching signal Sl2. is set. Further, when the character area is determined by the first area candidate string, the candidate string register 18 is
The contents of are passed through multiplexer 20 to provisional reference register 2.
Set to 1. When the search is completed, the character start signal Sl
4 is output, and further, the contents of the candidate column register 18 are transferred to the transfer register 19 by the signal Sl4,
Output data D2 is output from multiplexer 15 in accordance with the content D6.

The same applies to the next exploration area. FIG. 5 is a block diagram of a second embodiment of the present invention, in which character separation areas are determined by one search area.

In FIG. 5, the same parts as in FIG. 4 are given the same reference numerals, and 11 is a character buffer during search, 12 is a white column identification circuit, 13 is a normal reference timing signal generator, and 30 is a processing block. be. In this embodiment, the single character buffer and multiplexer 15 in FIG. 4 are eliminated, and the white row. The output signal S5 of the identification circuit 12 is also unified, and although the character area is not necessarily constant, it is possible to accurately determine the separation point between characters. As explained above, the present invention constantly searches the search area centered on the correct reference timing signal, and determines the character area using the provisional reference timing signal that corrects the error of the correct reference timing signal with respect to the character area. This has the advantage of being able to accurately determine character areas by correcting errors caused by paper expansion and contraction, typefaces that are out of character frame, etc., and can be used in handwritten optical character reading devices and the like.

[Brief explanation of the drawing]

1 to 3 are explanatory diagrams of the first embodiment of the present invention, and FIG.
This figure is a block diagram of the first embodiment of the present invention, and FIG. 5 is a block diagram of the second embodiment of the present invention. 1...Edge of paper, 2...Character frame, 3A,
3B, 3Z...Pulse of the positive reference timing signal, 4-1A, 4-1B...Pulse of the temporary reference timing signal at the initial time, 4-2A, 4-2B, 4-
3A, 43B...Pulse of updated temporary reference timing signal, 5-1, 5-2...Exploration area,
6-1, 6-2...Character area, 10...
・One character buffer, 11... Character buffer during search, 12, 12-1, 122... White column identification circuit, 13... Positive reference timing signal generator, 1
4... Control circuit, 15... Multiplexer, 16... Column counter, 17...
Multiplexer, 18... Candidate column register, 1
9... Determined sequence register, 20... Multiplexer, 21... Temporary reference register, 30...
...Processing block, S1...Sampling pulse signal, S2...Signal detecting paper edge, S3...Positive reference timing signal, S4...
...Temporary reference timing signal, S4-1...Temporary reference timing signal at initial time, S4-2, S4-
3...Updated temporary reference timing signal, S5
, S5-1, S5-2... White row identification signal, S6
...Signal indicating the exploration area, SlO...
Column count signal, Sll...temporary reference register update signal, Sl2...multiplexer 17 switching signal, Sl3...valid data set signal, S
l4...Character start signal, D1...Input character data, D2...Output data of multiplexer 15, D3...Input initial data of multiplexer 20, D4 ...Output of column counter 16,
D6: Output of decision string register 19.

Claims (1)

[Claims] 1. A white column identification signal is created in which a group of bits in a direction orthogonal to a character line in a character pattern signal is defined as a column, and a column in which the number of black bits in each column is equal to or less than a reference number is defined as a white column. , creating an exploration area indicating a predetermined range centered on a regular reference timing signal that is associated with a column at each character pitch interval from a predetermined reference point, and initially coinciding with the reference timing signal; Thereafter, a temporary reference timing signal that is sequentially updated is created, the white row identification signal is searched within each of the search areas, the white row closest to the temporary reference timing signal is detected, and the white row identification signal is detected at the position of this white row. A character area determination method characterized by separating a character area and updating and setting the next temporary reference timing at a position corresponding to one character pitch away from the position of the separated white row. 2 A white column identification signal is created in which a group of bits in the direction orthogonal to the character line in the character pattern signal is defined as a column, and a column in which the number of black bits in each column is less than or equal to a reference number is defined as a white column, and a white column identification signal is created based on a predetermined standard. A search area is created that shows a predetermined range from the point to the standard reference timing signal that is associated with a column at each character pitch interval. A reference timing signal is created, and the white row identification signal is set to two adjacent rows.
Search in two search areas, detect a pair of white columns at positions separated by one character pitch, separate the character area at the position of the pair of white columns closest to the temporary reference timing signal, and then A character area determination method characterized by updating and setting the next temporary reference timing at a position corresponding to one character pitch away from a column position.