JPH0797390B2 - Character recognition device - Google Patents

Description

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a character recognition device for recognizing printed characters and handwritten characters of newspapers, magazines, etc. and converting them into information amount such as JIS code.

2. Description of the Related Art A conventional character recognition device targets characters for vertical writing, horizontal writing, and a document with a clearly defined line spacing, character spacing, etc. Have been recognized. This limits the document targeted by the character recognition device, and in order to solve this problem, the line spacing such as vertical writing and horizontal writing is detected by detecting the line spacing using the two-dimensional Fourier transform of the input image. Extract
Even for documents of unknown format, a method of determining the order of recognition candidate characters has been adopted so that the meaning of the sentence can be understood. (For example, Hase and Hoshino “Periodic features of printed character strings”
(Problem theory (D), J65-D, 2, PP.298-299) Problems to be solved by the invention However, in the conventional technique of detecting the line direction of the input document by the two-dimensional Fourier transform of the input image, The processing takes a long time because the method that requires a large amount of calculation called the quadratic Fourier transform is used.

The present invention has been made in view of the above points, and detects a line direction of an input image by a simple method, and edits a recognition candidate character so that the meaning of a sentence can be understood even in a document whose format is unknown. It is an object of the present invention to provide a character recognition device that can be used.

Means for Solving the Problems In order to solve the above-mentioned problems, the present invention scans a pseudo-square (including a square) region in an image input by an image input unit in a vertical direction and a horizontal direction so that a character part is scanned. By obtaining the histogram of the pixels forming the pixel and comparing the average value of the character interval length defined by the number of continuous scanning lines in the character space where the value of the histogram is N pixels or less in the vertical and horizontal directions, A line direction determination unit for determining writing / horizontal writing is provided.

Effect According to the above-mentioned technical means, the present invention makes it possible to quickly extract the line direction of a document whose format is unknown and edit the recognition candidate characters so that the meaning of the sentence can be understood.

Examples Hereinafter, examples 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. An image input unit 1 scans an image including a character to be recognized, inputs the image with a binary signal, and stores it in the image memory 2. Reference numeral 3 denotes a row direction determination unit that sets a square area for the input image stored in the image memory 2 and scans the area to determine whether the input image is vertically or horizontally written. Reference numeral 4 denotes a character slicing unit, which scans the image memory 2 by using the line direction detected by the line direction determination unit 3 to detect a row address, scans the image memory 2 using the line address in units of lines, The character image to be recognized is cut out one by one using the projection of the unit image, and the character address on the input image of each character is detected. Reference numeral 5 denotes a recognition unit, which obtains a feature amount such as a stroke of the recognition target character cut out by the character cutout unit 4 and collates it with the feature amount of the character registered in the dictionary 6 in advance to identify the most similar character as a recognition candidate character. And An editing unit 7 uses the line direction obtained by the line direction determining unit 3 and the line address and the character address obtained by the character cutting unit 4,
The recognition candidate characters extracted by the recognition unit 5 are edited in an order in which the meaning of the sentence can be understood, and stored in the sentence memory 8 as a character code.

The character recognition device configured as described above will be described by taking the input image P shown in FIG. 2 as an example.

The image P input from the image input unit 1 includes a character portion 1 and a white portion O.
And is stored in the image memory 2 as binary data. First, the row direction determination unit 3 scans the input image P stored in the image memory 2 from the upper left by m pixels in the horizontal direction and m pixels in the vertical direction to form a character portion in a square area P ′ in the input image. The vertical histogram H _v and the horizontal histogram H _h of the selected pixels are as shown in FIG. In order to separate the character portion and the character portion from each other, for each of the histograms H _v and H _h, it is divided into a character portion whose histogram value is O pixels or less and a character portion having a pixel number larger than O pixels. Find the start address. Y _s1・ y _s2 …… y _si …… and x _s1・ x in Fig. 2
_s2 …… x _si …… is the start address of the character part, and y _e1・
y _e2 …… y _ei …… and x _e1・ x _e2 …… x _ei …… are the start addresses of the characters. From this address, the average value SP _Y of the vertical character spacing length in the square area P ′ is compared with the horizontal character spacing length SP _X. Where SP _Y = ave (Y _{s (i + 1)} -Y _ei ) SP _X = ave (X _{s (i + 1)} -X _ei ), where ave (f (i)) is i = 1, F in 2, ...
The average value of (i) is shown.

In the square area P ′, SP _Y > SP _X , and it can be seen that the square area P ′ and thus the input image P are written horizontally in the horizontal direction.

Next, in the character clipping unit 4, since the line direction of the input image P is determined to be horizontal writing, the vertical histogram H _l of the pixels forming the character part in the entire input image is obtained, and the line address of the input image P ( l _s1 , l _e1 ) ... (l _si , l _ei ) ... The results are shown in FIG. The row image L shown in FIG. 3 is extracted from the image memory 2 using the row address. When the histogram H _c is obtained by projecting the extracted row image L, it becomes as shown in FIG. The horizontal address (c _s1 , c _e1 ) ... (c _si , c _ei ) ... of each character is obtained from the histogram H _c , the character images to be recognized are cut out one by one, and the characters are combined with the line address. Determine the address.

The cutout rectangle R of the recognition target character is sequentially input to the recognition unit 5. FIG. 4 (a) shows the cut-out recognition target character "pine". For each of the cut out "pine" pixels, it is checked whether or not there are M or more (M is preset) including the pixel of interest in the direction indicated by the arrow in FIG. 4 (b), and the direction code is set. Strokes are extracted by checking the connectivity of each pixel for each direction code, and feature quantities such as the number, position, and length of strokes are extracted. FIG. 4A shows the stroke extraction result of the recognition target character "pine". The extracted feature amount is collated with the feature amount of the character registered in the dictionary 6, and the most similar character "pine" is set as the recognition candidate character.

The recognition candidate character of the recognition target character cut out from the input image P that is sequentially extracted from the recognition unit 5 is the editing unit 7 based on that the input image P is in horizontal writing and each character address.
"Matsu", "Lower", "Den", "Container" (registered trademark), etc. are arranged horizontally from the upper left to the lower right and edited so that the meaning of the sentence can be understood and stored in the sentence memory 8 with a character code. To be

With the character recognition device configured as described above, the meaning of the sentence can be understood by cutting out the recognition target character and further editing the recognition candidate character using the line direction and line address obtained by a simple method. You can create strings. Furthermore, a new document can be edited by connecting the text memory to a document processing device or the like.

In this embodiment, in the input image, the square direction of m × m pixels is scanned to determine the row direction. However, the pseudo square area having a difference of several pixels is scanned to determine the row direction. There is no problem.

EFFECTS OF THE INVENTION According to the present invention, in an image input by the image input unit, a pseudo square (including square) region is scanned in the vertical and horizontal directions to obtain a histogram of pixels forming a character portion, A simple method of determining vertical writing / horizontal writing by comparing the average value of the character interval length defined by the number of continuous scanning lines in the character space where the value of the histogram is N pixels or less in the vertical direction and the horizontal direction. By extracting the line direction of the input image at high speed by the method, it becomes possible to edit the recognition candidate characters so that the meaning of the sentence can be understood for a document whose format is unknown.

[Brief description of drawings]

FIG. 1 is a block diagram of a character recognition apparatus according to an embodiment of the present invention, FIG. 2 is an explanatory diagram of an input image and a method of determining a line direction and a line address of the input image, and FIG. 3 is a recognition target in the input image. FIG. 4 is an explanatory diagram of a character cutting method, and FIG. 4 is an explanatory diagram of a character recognition method. 1 ... Image input unit, 2 ... Image memory, 3 ... Line direction determination unit, 4 ... Character cutout unit, 5 ... Recognition unit, 6 ... Dictionary, 7 ... Editing unit, 8 ... Text memory .

Claims (1)

[Claims] 1. An image input section for inputting an image including a character to be recognized, and a pseudo-square (including a square) area in the image input by the image input section is scanned in the vertical and horizontal directions. Obtaining a histogram of pixels forming a character portion, and comparing the average value of the character interval length defined by the number of continuous scanning lines in the character portion where the value of the histogram is N pixels or less in the vertical direction and the horizontal direction. A line direction determination unit that determines vertical writing / horizontal writing by a character cutting unit that cuts out a recognition target character from an input image; a recognition unit that extracts a recognition candidate character by collating the recognition target character with a dictionary; A character recognition device having an editing unit for editing a candidate character.