JPH04372085A - Character reading method - Google Patents

Abstract

PURPOSE:To improve the state that characters become impossible in some cases, to be read by the difference of the position of the segment frame for each character according to the size of the width of a character which is placed at the left end when the conventional reading of a character string is performed by setting a segment frame at a position for each word to come in accordance with the upper end and the left end positions of a rectangle circumscribing the character string and comparing a character pattern within this frame with a registered character pattern. CONSTITUTION:An upper end position 12 for a character string 11 is determined from a X direction projection pattern 16 of a binary picture of the character string 11 and a center position 13 of a left end character is determined from a Y direction projection pattern 17 of the same character string 11. By using these two positions 12, 13 as references, each character frame 14 is set at a right position regardless of the width of the left end character. Further, for reading characters, a character circumscribing rectangle 15 is detected from the X direction projection pattern and the Y direction projection pattern of the characters within this frame 14 and the collation of the character pattern within this rectangle 15 with registered character patterns is performed.

Description

[Detailed description of the invention]

0001

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method of extracting an image of a character string as an electrical signal using a television camera or the like and reading each character of the character string from the signal.

0002

[Prior Art] Conventionally, this type of character reading method uses the upper and left edges of a rectangle circumscribing a character string to
A clipping frame is set at the position where each character should appear, a circumscribing rectangle of the character is cut out within each clipping frame, and the registered characters are matched against the character pattern within this character circumscribing rectangle to determine the character to be read. was.

0003

[Problems to be Solved by the Invention] However, in the above-mentioned character reading method, depending on whether the character that comes to the left end is a wide character, such as "4", or a narrow character, such as "1", The positions of the character extraction frames may be different, and adjacent characters may not be able to be separated. Additionally, the entire character may not be cut out correctly. Therefore, it is an object of the present invention to provide a character reading method that can eliminate the problem of difficulty in separating each character depending on the type of character that appears at the beginning of a character string, as described above.

0004

[Means for Solving the Problems] In order to solve the above-mentioned problems, the character reading method of the present invention provides a method for reading each character in a binary image of a character string (such as 11) arranged in the X direction or the Y direction. Set a cropping frame (character frame 14, etc.) at each position,
In a character reading method in which each character to be read is determined by comparing the character pattern to be read within this frame (within the character circumscribing rectangle 15, etc.) with a character pattern registered in advance (in the dictionary pattern memory 9, etc.), The top position (12, etc.) of this character string is determined from the X direction projection pattern of the character string, the center position (13, etc.) of this leftmost character is determined from the Y direction projection pattern of the character string, and this top position and leftmost character are calculated. The cutout frame is set using a point determined by the center position as a reference point.

[0005]

[Operation] In addition to detecting the top and left end positions of the character string, the center position of the character pattern separated at the left end position is determined from the vertical direction (Y direction) projection pattern of the character string, and the left end position of the character example is determined again. By doing this, even if characters with different widths appear at the left end, a character cutting frame is correctly generated for each character and reading is performed.

[0006]

[Embodiment] An embodiment of the present invention will be described below with reference to FIGS. 1 and 2. FIG. 2 is a block diagram showing a configuration as an embodiment of the present invention, and FIG. 1 is an explanatory diagram of the operation of FIG. 2 including a character string to be read. In FIG. 1, “2, 3,…, 7
” and “1, 2, ..., 6”, which are horizontally arranged in two rows, are the character string 11 to be read.
The configuration and operation of FIG. 2 will be explained together with reference to FIG. Reference numeral 1 denotes a TV camera as a photoelectric conversion sensor that images an object (in this case, a character string to be read), and raster-scans this imaged screen to output an analog image signal indicating the gray level of the object. 2 is an A/D converter that converts the grayscale image signal from the camera 1 into a binary digital signal (binarized signal) 2a
Convert to This binarized signal 2a is input to the image memory 7 and is stored as a binarized image of the character string 11.
It is input to the X projection section 3. The X projection unit 3 inputs this converted binary signal 2a and converts it into a character string 1 as shown in FIG.
A projection pattern 16 of 1 in the X direction is extracted. A microprocessor (also abbreviated as MPU) 8 reads out this projection pattern 16 from above and detects the upper end 12 of the character string as the position where the image first appears. Next, the Y projection unit 4 scans the image data of the character string stored in the image memory 7 in the Y direction to produce a Y image as shown in FIG.
A projection pattern 17 in the direction is extracted. And MPU8
reads out this projected pattern 17 in the Y direction from the left, and determines the center position (leftmost character center position) 13 in the X direction of the first cluster of patterns (that is, the projected pattern for the leftmost character). Further, in a similar manner, the center position of the block at the right end of the projection pattern is determined.

[0007] Next, the frame generation unit 5 extracts characters preset at the positions where each character should come, based on the upper end position 12 and leftmost character center position 13 of the character string detected in this way. A frame 14 is generated, the image memory 7 is scanned within this frame 14, and the image data within this frame 14 is outputted from the image memory 7. The X projection section 3 and the Y projection section 4 input this image data and extract projection patterns in the X and Y directions within the frame 14, and the MPU 8 reads out the X direction projection pattern and the Y direction projection pattern. The upper and lower ends of the X-direction projection pattern and the left and right ends of the Y-direction projection pattern are detected to obtain a circumscribed rectangle 15 of the character. Next, the matching processing unit 6 extracts the character circumscribing rectangle 15 from the image memory.
The character pattern cut out is read out, and a matching process is performed on this character pattern with a dictionary character pattern registered in the dictionary pattern memory 9 within the circumscribed rectangle to detect the degree of similarity. Based on the detected similarity, the character to be read is read as the dictionary character with the highest similarity.

[0008]

[Effects of the Invention] According to the present invention, X of a binary image of a character string
Find the upper end position 12 of the character string from the directional projection pattern, then find the center position 13 of the leftmost character from the Y direction projection pattern of the binary image of the character string, and use these positions 12 and 13 as reference positions to determine where the character will come from. A preset character frame 14 is set at the desired position, a character circumscribing rectangle 15 is detected within the frame, and the character to be read is read by matching with the registered dictionary character pattern. , even if the leftmost character is a large character or a small character,
Regardless of the type of character, a character cutting frame can be set at the correct position corresponding to each character, and each character can be accurately cut out and read by separating it from adjacent characters.

[Brief explanation of drawings]

[Fig. 1] An explanatory diagram of the present invention including characters to be read

FIG. 2 is a block diagram showing a configuration as an embodiment of the present invention.

[Explanation of symbols]

1 TV camera 2 A/D conversion unit 3 Center position 14 Character frame 15 Character circumscribing rectangle 16 X direction projection pattern 17 Y direction projection pattern

Claims (1)

[Claims] Claim 1: A cutting frame is set at each position of each character in a binarized image of character strings arranged in the X direction or the Y direction, and a character pattern to be read within this frame and a character pattern registered in advance are set. In a character reading method that determines each character to be read by comparing the characters, the upper end position of the character string is determined from the X-direction projection pattern of the character string, and the center position of this leftmost character is determined from the Y-direction projection pattern of the character string. , and the cutting frame is set using a point determined by the upper end position and the center position of the left end character as a reference point.