JPS63239569A - Character recognition device - Google Patents

Abstract

PURPOSE:To make rerecognizing operation speedy and accurate by extracting and rerecognizing a character area or character pattern corresponding to a candidate character for recognition when an error in character recognition occurs to a recognition part. CONSTITUTION:An image which includes a character area to be recognized is converted into a binary signal, which is stored in an image memory 2. A character segmentation part 3 segments a character string from the memory 2, further segments each character pattern to be recognized into a rectangle R, and stores it in a memory 4. The recognition part 5 finds the feature quantities of strokes of a character pattern in the rectangle R which is stored in the memory 4, and compares them with feature quantities in a dictionary 6 to extract the most similar character. If an error in recognition occurs to the recognition part 5, a rerecognition control part 7 re-sets parameters of the segmentation part 3 and parameters of the recognition part 5 and the results are displayed on a display part 8. Thus, the rerecognizing operation is made speedy and accurate.

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a character recognition device that recognizes and converts information into an amount of information such as a JIS code.

Conventional technology In conventional character recognition devices (Proceedings of the 1986 Institute of Electronics and Communication Engineers General National Conference Proceedings 1515), after cutting out characters in a character cutting section, the character recognition section compares the character pattern to be recognized with a recognition dictionary. The recognition part and complementary characters were extracted and output as recognition results.

Problems to be Solved by the Invention However, with the above configuration, if the recognition result is incorrect, the operator must specify a character pattern corresponding to the incorrectly recognized character to re-recognize the character pattern. There was a problem.

In view of this, the present invention extracts a recognition candidate character in a recognition unit, changes the extraction parameters and recognition parameters in order to perform re-recognition, and then uses a re-recognition control unit to extract a recognition target character area corresponding to the recognition candidate character and recognize it. It is an object of the present invention to provide a character recognition device that performs re-recognition by extracting a target character pattern, transferring it to a character segmentation unit and a recognition unit, performing character segmentation and character recognition again, and extracting recognition candidate characters.

Means for Solving the Problems In order to solve the above-mentioned problems, the character recognition device according to the present invention includes an image input section that inputs an image including a character region to be recognized, and a character recognition device that extracts images from the character region to be recognized according to preset cutout parameters. a character extraction unit that extracts a character pattern to be recognized; a recognition unit that extracts recognition candidate characters by comparing the recognition target character pattern with a dictionary according to preset recognition parameters; The recognition parameter of the recognition unit is reset, and the recognition target character area corresponding to the recognition candidate character obtained by the recognition unit is transferred to the character extraction unit according to the extraction parameter and the recognition parameter, and the character pattern to be re-recognized is extracted. A system that transfers the recognition target character pattern to the recognition unit and extracts re-recognition candidate characters, and recognizes the recognition target character pattern corresponding to the recognition candidate character obtained by the recognition unit according to the reset parameters and recognition parameters. The re-recognition control section has a system for extracting candidate characters for re-recognition.

Operation The present invention uses the above-mentioned technical means to change each parameter in order to perform re-recognition when the recognition candidate character obtained by the recognition unit is incorrect, and to extract the target to be re-recognized by the re-recognition control unit. There is no need for an operator to instruct and extract a recognition target character area corresponding to a recognition candidate character, and recognition work can be performed quickly and accurately.

Example Embodiments of the present invention will be described with reference to the drawings.

FIG. 1 is a block diagram of an embodiment of a character recognition device according to the present invention. Reference numeral 1 denotes an image input unit which scans an image including a character area to be recognized, inputs the image as a binary signal, and stores the image in the image memory 2. Reference numeral 3 denotes a character cutting unit which cuts out a character string from the binary image stored in the image memory 2, and further cuts out a rectangular character pattern to be recognized character by character.

Reference numeral 4 denotes a cutting memory, which stores the results obtained by the character cutting section 3. Reference numeral 5 denotes a recognition unit, which obtains feature quantities such as strokes of the character pattern to be recognized that exists inside the rectangle stored in the extraction memory 4, and compares it with the feature quantities of each character registered in advance in the dictionary 6. The most similar character is used as the recognition result. 7 is a re-recognition control unit, which resets the extraction parameters of the character extraction unit 3 and the recognition parameters of the recognition unit 5, and when the extraction parameters are changed, the recognition candidate characters already obtained by the recognition unit 5 are changed. The corresponding character region to be recognized is transferred to the character cutting section 3 to extract a character pattern to be re-recognized, and further the character pattern to be re-recognized is transferred to the recognition section 5 to extract candidate characters for re-recognition. Only recognition parameters are changed −
In this case, the recognition target character pattern corresponding to the recognition candidate character already obtained by the recognition unit 5 is transferred to the recognition unit 5, and the recognition candidate character is extracted again. A display section 8 displays the results of character extraction and character recognition.

The operation of the character recognition device of this embodiment configured as described above will be described below using an input image shown in FIG. 2 as an example.

An image as shown in FIG. 2 input from the image input section 1 is binarized and stored in the image memory 2, and the entire area of the input image is set in the character cutout section 3 as a character area to be recognized.

The character cutting unit 3 constructs a character string by projecting an image Q of an area set by the image input unit 1 or the re-recognition control unit 7 in the direction of the character string from among the input images stored in the image memory 2. Find the histogram H2 of the pixels, and find the start and end coordinates (ya++yε1) of the range where the value of the histogram H2 is one or more consecutive pixels (ya++yε1)...(y a + * y e
+ ) as the string coordinates and cut out the string image.

FIG. 3(a) shows a character string image. Next, a histogram H of pixels forming each character is obtained by projecting the character string image in a direction perpendicular to the character string.

Find the start and end coordinates (xs++xg+)・(Xa2
＋xl! ! ri @ a a (Xl191 XI!9
) and combine them with the character string coordinates to cut out the character pattern to be recognized into a rectangle R+ (i=1. . . +m). The extraction results are stored in the extraction memory 4 in Fig. 3 (b).
). Furthermore, the storage start address and storage end address of the rectangle R set in the cutout memory 4 are set in the recognition unit 5.

The recognition unit 5 focuses on the rectangle R1 included in the storage start address and the storage end address set by the character extraction unit 3 or the re-recognition control unit 7 among the rectangles R stored in the extraction memory 4. Regarding the character pattern P+ to be recognized surrounded by the target rectangle R1, it is checked whether or not there are M or more consecutive character patterns including the target pixel in the direction indicated by the arrow in FIG. 4(b), and a direction code is assigned to the target pixel. , examine the connectivity of pixels for each direction code, extract strokes, and calculate the number, position, length, etc. of strokes using n-dimensional feature quantities fii
(J=1....,n). Figure 4 (
Figure a) shows the stroke extraction results for the character pattern r to be recognized. Furthermore, the obtained recognition target character pattern P1
The distance D i k between the feature amount f1 of
The one with the smaller ik is set as the recognition result A1.

The re-recognition control unit 7 resets the cut-out parameters used for re-recognition in the character cut-out unit 3, and resets the recognition parameters in the recognition unit 5. FIG. 5 shows an example of the cutting parameters and recognition parameters to be reset. If the extraction parameters are changed, both extraction and recognition are re-executed regardless of whether or not the recognition parameters have been changed.
If the extraction parameters are not changed and only the recognition parameters are changed, only recognition is re-executed. In this embodiment, a case will be described in which the extraction parameter is changed from full-width to half-width and the recognition parameter is changed from kana-kanji to numbers, so both extraction and recognition are re-executed. First, the recognition unit 5 extracts a re-recognized character area corresponding to the recognition candidate character already obtained. The method will be explained with reference to FIG. 2nd and 3rd characters of recognition candidate characters
Select characters and extract the character area of those characters. The character pattern to be recognized corresponding to the second character is rectangle R2, and the character pattern corresponding to the third character is rectangle R3, so the character area including rectangles R2 and R3 has the coordinates (X 112 ,
Y I! +) and the coordinates (XE3.Yell), and this area is extracted as the re-recognized character area. The extracted re-recognized character area is set in the character segmentation section 3 and character segmentation is performed. The rectangle of the re-recognized character pattern cut out by the character cutout section 3 is newly stored in the cutout memory 4, and the storage start address and storage end address of the rectangle are set in the recognition section 5 so as to indicate the cutout rectangle R1. Next, the recognition unit 5 performs re-recognition to obtain re-recognition candidate characters.

Similarly, it is also possible to re-execute only the recognition.

The display unit 8 displays the binary image stored in the image memory 2, the rectangle R8 obtained by the character cutting unit 3, and the recognition result A obtained by the recognition unit 7. Figure 6 shows an example of displaying the recognition candidate characters extracted by recognition and the input image, and Figure 7 shows an example of displaying the re-recognition candidate characters extracted by re-recognition and the input image. show.

Effects of the Invention In the present invention, when a character recognition error occurs in the recognition unit,
The character area or character pattern corresponding to the recognition candidate character is extracted and re-recognized. This eliminates the need for the operator to search for and re-recognize the character area or character pattern that corresponds to the recognition candidate character (and prevents the operator from accidentally specifying a character area or character pattern that does not correspond to the recognition candidate character. When only recognition parameters are changed, unnecessary re-extraction is not performed and only re-recognition is performed.In other words, re-recognition can be performed quickly and accurately.Therefore, the character recognition device according to the present invention , is of extremely high industrial value.

[Brief explanation of the drawing]

Fig. 1 is a block diagram of a character recognition device in an embodiment of the present invention, Fig. 2 is an explanatory diagram showing an example of an input image, and Fig. 3 is a character cutting method in an embodiment of the present invention. FIG. 4 is an explanatory diagram showing a feature extraction method in an embodiment of the present invention, FIG. 5 is an explanatory diagram of extraction parameters and recognition parameters in an embodiment of the present invention, FIG. 6 is an explanatory diagram showing a method for extracting a recognized character area corresponding to a recognition candidate character in an embodiment of the present invention, and FIG. FIG. 3 is an explanatory diagram of a display unit displaying re-recognition candidate characters and an input image. 1... Image input section, 2... Image memory section. 3... Character cutting section, 4... Cutting memory, 5
... Recognition unit, 6... Dictionary, 7... Re-recognition control unit,
8...Display section. Name of agent: Patent attorney Toshio Nakao and one other person

Claims (1)

[Claims] an image input unit for inputting an image including a character region to be recognized;
A character extraction unit that extracts a character pattern to be recognized from a recognition target character area according to preset extraction parameters, and a recognition unit that extracts recognition candidate characters by comparing the recognition target character pattern with a dictionary according to preset recognition parameters. section, reset the extraction parameters of the character extraction section and the recognition parameters of the recognition section, and set the recognition target character area corresponding to the recognition candidate character obtained by the recognition section to the character extraction section according to the extraction parameters and the recognition parameters. A system for transferring and extracting a re-recognition target character pattern, transferring the re-recognition target character pattern to the recognition unit and extracting re-recognition candidate characters, and recognition obtained by the recognition unit according to the reset extraction parameters and recognition parameters. A character recognition device comprising a re-recognition control section having a system for transferring a recognition target character pattern corresponding to a candidate character to the recognition section and extracting re-recognition candidate characters.