JPH04271488A - System for detecting noise - Google Patents

Abstract

PURPOSE:To detect even the black picture element block of longitudinal or horizontal shape when it is outside the reference value range, as a noise and to improve the accuracy of the noise detection by performing the line classifica tion of the black picture element block as a means to detect the noise, obtaining standard height and standard width for each line, specifying the range on the basis of it and detecting the noise. CONSTITUTION:Black picture element block information obtained from a character segmenting circuit used by image processing is used and the black picture element block is classified into plural lines. Next, the standard width value and standard height value, which become the reference to detect the noise for every line, are obtained. 2, 4 and 6, in which the standard height value and standard width value cannot be decided, are considered to be noise lines and the information is removed. Concerning 1, 3 and 5 lines in which the standard height value and standard width value are decided, the standard height value and the black picture element block height, and the standard width value and black picture element block width are compared for each line, and the black picture element block of 7, 8 and 9 judged to be the noise is removed. As this result, only the noise can be removed.

Description

[Detailed description of the invention]

0001

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a character recognition device, and more particularly to noise detection processing in preprocessing.

0002

[Prior Art] Conventionally, noise detection in this type of device uses height information and width information of black pixel block information, checks whether the values of the height information and width information are within an allowable range, and then If it is outside the range (below α or more than β), this black pixel cluster is regarded as noise. Note that α and β are the lower and upper limits of the allowable range.

0003

[Problems to be Solved by the Invention] In the conventional method described above, only when both the height information and the width information of a black pixel block are outside the allowable range, it is considered as noise. If either one satisfied the conditions, it was not considered noise. Therefore, if the shape of the black pixel block is horizontally long or vertically long, it cannot be detected as noise even if it is actually noise, and the accuracy of noise detection is reduced. Also, if we add conditions to judge these as noise, for example, “1
", "-", etc. were not added as conditions because they might be interpreted as noise.

An object of the present invention is to provide a noise detection method capable of highly accurate noise detection in image processing of a character recognition device.

[0005]

[Means for Solving the Problems] The noise detection method of the present invention is a character recognition device that recognizes printed characters printed on paper sheets, and detects the black color obtained by scanning the paper surface and using a character cutting circuit. A noise detection method that detects noise using pixel block information includes means for detecting lines using the black pixel block information and classifying the lines into black pixel block rows, and using the line-classified black pixel block information. and a means for detecting noise using the calculated standard height and standard width.

[0006]

DESCRIPTION OF THE PREFERRED EMBODIMENTS Next, the present invention will be explained with reference to the drawings.

FIG. 1 is a block diagram showing the operating procedure of an embodiment of the present invention. First, a sheet of paper is scanned by an image input device, and the image information (S1) outputted from the image input device is subjected to character cutting out using a character cutting circuit (S2). FIG. 2A shows black pixel block information output from the character extraction circuit as a result of character extraction. Black pixel blocks are classified into lines using this black pixel block information. Line classification is to classify black pixel blocks into rows.

Next, line classification processing will be described. Let an arbitrary black pixel block a be the black pixel block information Xa, Ya, Ha
, Wa (see Figure 2(B)). Let b be any black pixel block other than black pixel block a, and let the black pixel block information be Xb.
, Yb, Hb, Wb. Using the black pixel block a as a reference, a comparison process is performed with the black pixel block b based on conditional expression -1.

Conditional Expression-1 Ya≦Yb +Hb /2≦Ya +Ha or Yb≦Ya +Ha /2≦Yb +HbIf this conditional expression is satisfied, black pixel block a and black pixel block b are considered to be on the same line. Line classified.

Next, using black pixel block b as a reference black pixel block,
A comparison process is performed by applying conditional expression-1 to black pixel clusters a and any black pixel clusters other than black pixel clusters b that have not yet been line-classified. This process is repeated, and when no black pixel block satisfying the conditions is detected, the classification process for one line ends. The above processing is repeated until all the black pixel blocks output from the extraction circuit are classified into lines. The results of the line classification process are shown in FIG. 2(C). The process described above is the line classification process, which corresponds to step S3 in FIG.

Next, the standard height value and standard width value of the black pixel block are obtained for each line. The standard height value is the average value of the height of black pixel clusters that appear to be characters, and the average value of the height information of the black pixel cluster information that satisfies conditional expression-2 is taken as the standard height value.

Conditional expression-2 α≦height value≦β and width value (±γ)=height value α, β
are the minimum height value and maximum height value of the print character, and the optimum values for normal print characters are α=2 (mm) and β=7 (mm). γ is the margin value, which is 0.7 (
mm) is the optimum value.

The standard width value is the average value of black pixel block values that appear to be characters, and the standard width value is the average value of the width information of black pixel block information that satisfies conditional expression-3.

Conditional expression-3 α≦width value≦β and height value (±γ)=width value α, β,
γ is the same value as used in Conditional Expression-2. Note that lines for which either the standard height value or the standard width value could not be determined are considered to be noise lines and are deleted. (
(See Figure 3(A)). The process described above is the line classification process and corresponds to step S4 in FIG.

After the standard height value and standard value for each line are determined, noise is detected by comparing the line-classified black pixel block information with the standard height value and standard width value of that line. . In the noise detection process, the height information and the standard height value of the black pixel block information, and the width information and the standard width value are expressed in conditional expression-4.
Comparison processing is performed based on the above, and if a black pixel cluster that does not satisfy the conditions is detected, the black pixel cluster is regarded as noise and removed.

Conditional expression-4 Standard height/2-μ≦Black pixel block height≦Standard height/2+μ and Standard width-μ≦Black pixel block width≦Standard width+μ or Standard width/2-μ≦ Black pixel block width ≦ standard width / 2 + μ and standard height - μ ≦ black pixel block height ≦ standard height + μ or, standard width - μ ≦ black pixel block width ≦ standard width + μ and standard height - μ ≦ Black pixel block height≦standard height+μ where μ is the same as the value used in conditional expression-2. The results of the noise removal process are shown in FIG. 3(B). The process described above is the noise detection process, and corresponds to step S5 in FIG. By performing noise detection and removal in this manner, it is possible to overcome the conventional drawbacks and improve the performance of noise detection.

[0017]

[Effects of the Invention] As explained above, the present invention classifies black pixel clusters into lines as a means of detecting noise, determines the standard height and standard width for each line, and uses them as a reference to define a range and detect noise. Therefore, even if it is a black pixel block that is vertically or horizontally long, it can be detected as noise if it is outside the standard value range, and "1" or "-" are not considered noise because they are within the standard value range. This improves the accuracy of noise detection.

[Brief explanation of the drawing]

FIG. 1 is a diagram showing an operation procedure of an embodiment of the present invention.

FIG. 2 is a diagram showing an example of black pixel block information, in which (A) is black pixel block information obtained from an extraction circuit, (B) is black pixel block information of an arbitrary black pixel block a, and (C) is a diagram showing an example of black pixel block information. 1 and 2 show examples of line classification of black pixel block information.

FIG. 3 shows line classification results of black pixel block information, (A)
(B) shows the division between the black pixel block information line and the noise line, and (B) shows the noise contained in the black pixel block after removing the noise line in (A).

Claims (1)

[Claims] Claim 1: Noise detection in a character recognition device that recognizes printed characters printed on paper sheets, which scans the paper surface and detects noise using black pixel block information obtained from a character cutting circuit. The method includes means for detecting lines using the black pixel block information and classifying the lines into black pixel blocks row by line, and means for detecting lines using the black pixel block information and classifying the lines into black pixel blocks row by line, and determining the standard height of the black pixel block for each line using the line classified black pixel block information. and means for calculating a standard width; and means for detecting noise using the calculated standard height and standard width.