JPH04167084A - Character reader - Google Patents

Abstract

PURPOSE:To detect stable character lines at all times and to detect and segment characters by scanning a character image in the same direction with and at right angles to the character lines and extracting maximum density value data on scanning lines, and detecting the characters from their density differences. CONSTITUTION:The character image stored in a many-valued image memory 13 is scanned in the character entry direction (character line direction) first to extract the maximum density value data A1 on each scanning line, and this extracted maximum density value data A1 is the largest density value data among density value data on each scanning line. Differential data A2 is found from the maximum density value data A1 and the character lines are detected and segmented (a1 - a2) according to the maximum value of the differential data A2. Then the character image on the detected character lines (a1 - a2) is scanned at right angles to the character lines to extract maximum density value data B1 on the respective scanning lines and the characters are detected and segmented (b1 - b2, b3 - b4, and b5 - b6) from the maximum value among differential data B2 found from the data B1. Consequently, the character lines are stably detected at all times and the respective characters are stably detected and segmented.

Description

[Detailed Description of the Invention] [Object of the Invention] (Industrial Application Field) The present invention relates to a character reading device that optically reads characters, etc. on a recording medium. This invention relates to character detection and extraction technology for detecting and extracting characters.

(Prior art) In this type of character reading device, after converting an input multivalued character image into a binary character image, character lines are detected, and then each character is detected and extracted. It is common to do so. However, with this method, if the density level is locally low, the binarized image will be missing, and especially if the area around the character is missing, the center of the character will shift, which will affect character recognition.

On the other hand, in order to avoid these problems, it is possible to extract multiple binarized images, perform character recognition, and use the one with the best recognition result as the final result, or divide the image into multiple parts and perform the binarization process in units of divisions. Although some methods are designed to obtain binarized images, they all have the drawback of requiring a large circuit scale.

(Problem to be Solved by the Invention) As described above, when character lines are detected and each character is detected and extracted after converting an input multi-valued character image into a binary character image, it is difficult to If the surrounding density level is low, the binarized image will be missing, the center of the character will shift, and character recognition will be affected. On the other hand, a large circuit scale is required to implement a means for preventing image deletion.

Therefore, the present invention provides the following advantages:
To provide a character reading device that can always stably detect a character line and detect and cut out each character, and can realize this with a relatively simple configuration.

[Structure of the Invention] (Means for Solving the Problems) The character reading device of the present invention repeatedly scans an input multivalued character image in the same direction as the character line, thereby reading the inputted multi-valued character image on each scanning line. A character line is detected from the density difference between the first density value extraction means for extracting the maximum density value data or the minimum density value data, and the maximum density value data or the minimum density value data extracted by the first density value extraction means. By repeatedly scanning the character image within the character line detected by the character line detection means in the direction perpendicular to the character line, maximum density value data or minimum density value on each scanning line is determined. a second density value extraction means for extracting data; a character detection means for detecting a character from the density difference between the maximum density value data or the minimum density value data extracted by the second density value extraction means; and character recognition means for recognizing each character detected by the means.

(Function) The present invention does not detect a character line or each character in a binary character image as in the past, but detects a character line in the same direction as the character line for an input multivalued character image. Maximum (or minimum) 1 displacement data on each scanning line is extracted, and a character line is detected from each density difference of the maximum (or minimum) S displacement data. Then, for the multivalued character image within the detected character line, the maximum (or minimum) density value data on each scanning line in the vertical direction to the character line is extracted, and each of the maximum (or minimum) density value data is It detects characters based on density differences.

Note that when the density level of the character image is higher than the density level of the background, the maximum density value data is used, and when the density level of the character image is lower than the density level of the background, the minimum density value data is used.

(Example) Hereinafter, an example of the present invention will be described with reference to the drawings.

In FIG. 1, the photoelectric conversion unit 1 includes, for example, a light source that irradiates light onto a recording medium on which characters or the like are recorded, and a CCD type sensor that receives reflected light from the recording medium and converts it into an electrical signal. It consists of a photoelectric converter, an optical system that guides light from a light source onto a recording medium, and guides the reflected light to the photoelectric converter, and converts it into an electrical signal (analog signal) by scanning the recording medium with light. It looks like this. The analog signal from the photoelectric converter 1 is converted into a multi-value digital signal by the A/D converter 2 and stored in the multi-value image memory 3 as a character image. A maximum/minimum density value extraction section 4 extracts maximum or minimum density value data from the character image stored in the multivalued image memory 3. Then, in the character detection and extraction section 5, detection and extraction of each character is performed using the maximum or minimum density value data extracted by the maximum/minimum density value extraction section 4, and then the character recognition section 6 performs detection and extraction for each character. Each character is recognized.

FIG. 2 shows a schematic diagram of the maximum/minimum density value extracting section 4, and the processing using maximum density value data will be described below as an example. The character image input from the multilevel image memory 3 is sent to the input terminal P of the comparator 11 and the first tristate 12.
is input. On the other hand, the maximum/minimum concentration value storage memory 13
The maximum concentration value data already stored in the pulse PI
(See FIG. 3) is input to the input terminal Q of the comparator 11 via the second tri-state 14 which is turned on and off according to FIG. 3, and the data input to both input terminals P and Q are compared.

At this time, if the maximum/minimum switching signal is set to the mode in which the maximum value is detected, when the value at the input terminal P is larger than the value at the input terminal Q, the output is used to detect the maximum value. The flip-flop 18 is set. Conversely, if the maximum/minimum switching signal is set to the mode that detects the minimum value, when the value of the input terminal P is smaller than the value of the input terminal Q, the output of the NOR gate 16,
Flip-flop 18 is set via AND gate 17.

Next, the set output of the flip-flop 18 is the pulse P2
(see FIG. 3) is extracted by the Nandt gate 19, and its pulse P2 becomes the write pulse WT and is applied to the maximum/minimum density value storage memory 13. The density value data that has passed through the tristate 12 is stored in the maximum/minimum density value storage memory 13 as new maximum density value data.

The above operations are performed for each scanning line using the time-division pulses PI and P2. The counter 20 generates an address for the maximum/minimum density value storage memory 13, and is incremented every time one scan is completed in response to a scan end signal.

FIG. 4 shows an example of character detection and extraction using these processes. The character image stored in the multivalued image memory 13 is first scanned in the character writing direction (character line direction),
Maximum density value data A1 on each scanning line is extracted. As the maximum density value data A1, the maximum density value data among the density value data on each scanning line is extracted. In FIG. 4, maximum density value data A on each extracted scanning line is shown.
Differential data A2 is obtained from the differential data A2, and character lines are detected and cut out (al to a2) from the maximum value of the differential data A2. Next, the character image within the detected character line (al to g2) is scanned in a direction perpendicular to the character line, and maximum density value data B1 on each scanning line is extracted. Next, differential data B2 is obtained from the extracted maximum density value data B1 on each scanning line, and character detection cutting (b
l~b2. b3-b4. Perform b5-be). Then, detection cutting out (bl-b2, b3~b4.b5~be
) is subjected to similar processing in the direction of the character line, the coordinates of the circumscribed rectangle of the character are determined, and character detection and extraction is completed. Thereafter, character recognition is performed by binarizing the character image within each detected and extracted character.

Although the processing using the maximum density value data has been described above, the same idea can be applied to the processing using the minimum density value data.

In addition, in the above embodiment, character detection and extraction is performed by obtaining differential data from maximum (minimum) density value data, but a histogram created from maximum (minimum) density value data using maximum (minimum) density value data It is also possible to detect and cut out characters by binarizing them using the threshold value determined in .

[Effects of the Invention] As described in detail above, the character reading device of the present invention does not detect character lines or individual characters using a binary character image as in the past, but detects input multi-valued characters. For the character image of
The maximum (or minimum) density value data on each scanning line in the same direction as the character line is extracted, and the character line is detected from the density difference between the maximum (or minimum) density value data. Then, for the multivalued character image within the detected character line, the maximum (or minimum) density value data on each scanning line in the vertical direction to the character line is extracted, and each of the maximum (or minimum) density value data is It detects characters based on density differences.

Therefore, according to the present invention, it is possible to always stably detect character lines and detect and extract each character without affecting character recognition as in the past, and this can be achieved with a relatively simple configuration. We can provide a character reading device that can.

[Brief explanation of drawings]

The figures show one embodiment of the present invention; FIG. 1 is a block diagram schematically showing the overall configuration, FIG. 2 is a block diagram schematically showing the configuration of the maximum/minimum density value extraction section, Figure 3 is a diagram showing the pulses used in the maximum/minimum density value extraction section;
The figure is a diagram for explaining an example of extracting maximum density value data and an example of detection cutting out. 1... Photoelectric conversion section, 2... A/D conversion section, 3...
Multivalued image memory, 4... Maximum/minimum density value extraction section, 5
...Character detection cutting section, 6...Character recognition section, 13...
・Maximum/minimum concentration value storage memory. Applicant Representative Patent Attorney Takehiko Suzue Ward 2 #g3 Ward

Claims (2)

[Claims] (1) A first density value extraction means for extracting maximum density value data or minimum density value data on each scanning line by repeatedly scanning an input multivalued character image in the same direction as the character line; , a character line detection means for detecting a character line from the density difference between the maximum density value data or the minimum density value data extracted by the first density value extraction means; a second density value extraction means for extracting maximum density value data or minimum density value data on each scanning line by repeatedly scanning the character image in a direction perpendicular to the character line; and a character recognition means for recognizing each character detected by the character detection means. Characteristic character reading device. (2) The first and second density value extraction means extract maximum density value data when the density level of the character image is higher than the density level of the background, and the density level of the character image is higher than the density level of the background. 2. The character reading device according to claim 1, wherein minimum density value data is extracted when the density is low.