JPS6089284A - System for cutting out character line - Google Patents

Abstract

PURPOSE:To prevent omissions of an upper part of a character first by cutting out a character line of a specific line, which is a heading line, preceding to a line whose horizontal projective value exceeds a threshold value. CONSTITUTION:A CPU11 examines a character cutting out condition STATUS when receiving an interruption signal, and returns to a program of a control height Ln=0 when STATUS is equal to 0. When STATUS is equal to one and when a count value of a horizontal projection counter 20 is more than a specified threshold value, the CPU11 makes either an image memory 24 or 25 store the data of one line buffer 22 against an I/O controller 12 and adds 1 the data to the Ln. Next, when STATUS is equal to two, the CPU11 controls 12 to store the image data of the buffer 22 to a memory 24 or 25 by making a line directly before the character line, whose horizontal projective value first exceeds a threshold value, a heading line of the characteristic line, to store said image data and then cuts out character line.

Description

DETAILED DESCRIPTION OF THE INVENTION [Technical Field] The present invention relates to a character f1 extraction method in a single character recognition device or the like.

[Prior art]

The conventional method of cutting out the character f77 will be explained with reference to FIG.

The image data subjected to noise removal is input to the input terminal l, and is temporarily stored in the l line buffer 2. Horizontal projection counter: 3 counts "'1" bits (black image 1f4) of image data for each line. When the input of image data for each line is completed, the control unit 4 compares the count value of the horizontal projection counter 3, that is, the horizontal projection value for each line, with a specific threshold value. When the horizontal projection value exceeds the threshold value for the first time in a certain line, the control section 4 stores the image data of a predetermined number of lines after that line in the image memory 5 via the one-line buffer 2.

In this way, conventionally, the line whose horizontal projection value first exceeds the threshold value is taken as the first line of a character line, and a predetermined number of lines of image data are cut out from that line.

Now, the image data read by scanning a manuscript etc. is
Even if noise such as isolated points is removed, 97 emails still contain a considerable amount of noise due to dirt on the background of the original, etc. In order to avoid erroneous detection of the first line of character lines due to such noise, The above threshold value cannot be set too small. As a result, in the case of a character line where the horizontal projection value of the first line is small, the line after the true first line is mistakenly detected as the first line and the character line is cut out, resulting in the upper edge of the character image being missing and character recognition. It has drawbacks such as causing errors.

〔the purpose〕

SUMMARY OF THE INVENTION An object of the present invention is to provide a method for cutting out double lines that eliminates the above-mentioned drawbacks.

〔Example〕

Embodiments of the present invention will be described in detail with reference to FIGS. 2 to 4.

FIG. 2 is a block diagram with some parts of the character recognition device omitted. ■1 is CPU, 12 is I, /1. ) 21 is a controller, 13 is a scanner, and 14 is a noise removal unit.

] 5 and 16 are 1 line buffers, 17 is a (×3 register matrix, 18 is a ROM, 19th scanner control section, 20 is a horizontal projection counter, 21 and 22 are 1 line buffers, 23 and 26 are switch circuits, 24 and 25- is an image memory.

Note that each part of the noise removal unit 14 is
, 22, and the control signals output from the I10 controller 12 to the image memory 24.2F'i are omitted in the figure.

FIG. 3 is a schematic flowchart 1- of a program executed by the CPU II, and FIG. 4 is a flowchart of interrupt processing by the CPU 11.

The operation of the apparatus of this embodiment will be explained with reference to these flowcharts.

The CPU II instructs the scanner control unit 19 to initialize the scanner 13 through the I10 controller 1 to the roller 12 (step 101), and resets the control variables to initial values (step 102). As a control variable, the control height L
n, line spacing height r-n G A P, character line cutting shape M S
There is T A T U S and L n = O5L n
G A P = 01 STATUS = set to O
Sign. Then, it is monitored whether the original has been sent to the scanner 13 (step 103).

When the original is loaded, the CPU II enters the STA TU
Set S=2 (step 104). The image data input serially from the scanner 13 is shifted to the right through the lowest registers G, H, and I of the 3×3 register matrix 17, and is delayed by one line by the l-line buffer 16. The middle registers D, E, and F of the 3×3 register matrix 17 are shifted to the right, and
After AI is further delayed by one line by buffer 15,
, B, and C are shifted to the right by 1-. In this way, the data of 3 pixels of 3 consecutive lines, 9 cars in total, are stored in 3x3 registers 1 to 17 (44), :,
It is given to the address input of ROM1.8. ROM2S
The method removes isolated black or white pixels as noise, and outputs the data of the pixel of interest (Δ) that has been subjected to noise removal processing. The image data that has been subjected to L re-noise removal in the ROM 18 is sequentially stored in the 1-line buffer 21, and the image data that is output from this 1-line buffer 21 with a delay of 1 line is sequentially stored in the line buffer 22 at the next stage.
It is multiplied by j. The horizontal projection counter 20 counts "1" bits (black pixels) of image data for each line.

The operation of this horizontal projection counter 20 is controlled by a 1° buffer 21.

Character line cutting is performed in the interrupt processing 111j shown in FIG. Each time the scanner 13 reads one line of the original, a synchronization signal (S'YNC) is output from the scanner control section 19, and this synchronization signal is used as an interrupt signal in the I10 controller 1-c.
It is input to the CPU 11 via the l-ra 12. cputt
When it receives an interrupt signal, it checks S T A T U S (step + + 1), and 5TA1' U S =
If it is 0, return to the program with L n = O (step 112).

If 5TATUS=1, it is determined whether the count 1 of the horizontal projection counter 20 (the horizontal projection value of the line whose image data is stored in the 1-line buffer 21) is greater than or equal to a predetermined threshold (step 113). If Y'ES,
CPtJl,] transfers the image data of 1 line buffer 22 to I10 controller 1-roller 1 to image memory 2.
4 or to be stored in the image memory 25, and 1 is added to Ln (step 114). ij7
The image memory 24 and the image memory 25 are used alternately for cutting out character lines and cutting out characters.
When connected to the fourth side, the image memory 25 is selected by the switch circuit 26.

Such switching control of the switch circuit 23 and the switch circuit 26 is performed by the T10 controller 1-roller 12.

If the determination result in step 113 is No, it is determined whether Ln is greater than or equal to the minimum height of a character line (step 115)-
If YES, it is determined that the I3IJ output of one character line has been completed, and the storage of the image data in the image memory 24 or the image memory 25 is stopped, and the switch circuit 2
The I10 controller 12 is commanded to switch 3.26, and the S1A TU S is set to O (step If(i)).

If the determination result of the switch 115 is NO, the controller 1 to roller 121 issues an IH command to continue storing the image data of 22 into the image memory 2/or the image memory 25; L, n
1 is added to (step 1j7).

If S T A T U S = 2, step 11
The same determination as in step 3 is made (step 118). Judgment viscosity 1) 1
If is NFS, 1 line buffer 22 and image 5-
The line in which the data is accumulated (the line immediately before the line whose horizontal projection value first exceeds the threshold value) is determined to be the first line of the character line, and the image data in the line buffer 22 is transferred to the image memory 2/I or the image memory. 25, the T10 controller 1-roller 12 is instructed to store the control variable in S
T A T U S = ], set Ln=1 (
Step 119). If the judgment result is No, L n G
Add 1 to AP (step) 20).

Again in FIG. 3, CI) LJ I + determines whether reading of the original is complete (step +05), and if it is not completed, S
Step 1o6 to determine whether T A 1” US is 0
), if YES, set S T A T U S to 2 (step 107), and perform character extraction (step 108).
).

When character extraction is completed, the process returns to step 105.

In the character cutting process, the CPU I scans the image memory 24 or the image memory 25 (the side selected by the switch circuit 26) via the 1-line buffer 22, finds the vertical projection of the character line, and cuts the characters at the breaks in the vertical projection. is cut out and transferred via the I10 controller 12 to a character recognition unit that generates image data for each character.

If the determination in step 105 is YES, it is determined whether the character extraction/transfer processing for the last character line extracted has been completed (step 1o9). If Yl:S, scanner 13
Post-processing such as stopping , sending 1·jl i' to the character recognition unit, etc. is performed (step I11.).

In this embodiment, a 1-line buffer is first used to save the image data of the line immediately before the line where the horizontal projection value reaches the threshold starboard. :! rj I am tracking, but in the state of S'I'ATUS = 2 (extracting between lines), the image 111-7 in the 1-line buffer 21 is transferred to the same line in the image memory 24 or the image memory 25. 1'') run buffer 22 may be omitted by overriding the line 1 and sequentially storing data from the second line and subsequent lines when S TΔ'l'U S =1.

In addition, in this embodiment, a character line is cut out by using the line immediately before the line whose horizontal projection value exceeds the threshold 4fi as the first line of the character line, but if you move forward by 1°, the line immediately before the line is cut out. It is also possible to set it as the first line.

〔effect〕

As detailed above, according to the present invention, a character line is first cut out using a specific line before the line whose horizontal projection value exceeds the threshold as the first line, so a high threshold that is not affected by noise can be set. Even if you select this option, problems such as the true first part of a character line disappearing and the second end of a character being missing can be solved.

[Brief explanation of the drawing]

Figure 1 is a block diagram showing the conventional character line extraction method;
The figure is a partially omitted block diagram of the character recognition device according to the present invention, Figure 3 is a schematic flowchart of the CPU program in Figure 2, and Figure 4 is a flowchart 1- of the interrupt processing of the same CP tJ. . 11... CPU, 12... I10 controller,
13... Scanner, 14... Noise removal unit, 19
...Scanner control unit, 20...Horizontal projection unit counter, 21.2'2...1 line buffer, 23.26...
・Switch circuit, 24.25 image memory. Representative Patent Attorney Makoto Suzuki A'1 Figure O/2 Figure 4

Claims (1)

[Claims] (1) In the character line extraction method that cuts out the image data of a character line based on the horizontal projection of image data, a specific line before the first ``u'' whose horizontal projection value exceeds a threshold value is defined as the first line of the character line. A character line cutting method that is characterized by the following.