JPH02253383A - Picture processor - Google Patents

Abstract

PURPOSE:To automatically extract and classify the areas of a character string, a graphic, a photograph, a picture, a line segment, and a noise in a picture by providing a rectangular area coordinate extracting part, a rectangular area characteristic extracting part, and a rectangular area classifying part. CONSTITUTION:A rectangular area is extracted from the picture by a rectangular area coordinate extracting part 3, the characteristic of the extracted rectangular area is extracted by a rectangular area characteristic extracting part 4, and the extracted characteristic is compared with the characteristics of the character string, graphic, photograph, line segment and noise prepared beforehand, and to which of the character string, graphic, photograph, line segment or noise the rectangular area corresponds is decided by a rectangular area classifying part 5. By using the method, every one character is segmented and recognized by character recognizing technique, respectively inherent processings are executed for the graphic, photograph, line segment and noise areas, and thus the input picture can be more flexibly processed.

Description

[Detailed description of the invention] Industrial applications Character strings from documents in unspecified formats such as newspapers and magazines.

The present invention relates to an image processing device that extracts and classifies regions of charts, photographs, line segments, and noise.

Conventional technical image processing devices that separate character areas and seven graphic areas from an image containing a mixture of one character and one figure include those that display the input image and allow an operator to specify it using a mouse, and those that automatically specify the area without operator intervention. There is something to do. Automatic pattern recognition without operator intervention involves forcibly cutting out the entire image into small areas for each character based on a predetermined format without distinguishing between characters and figures. technology is used to recognize each character, the recognition results of the small areas for each character are determined whether or not they can be recognized as characters, and the continuity of the small areas for each character is examined using the determination results. Character areas and graphic areas were classified (for example, Japanese Patent Laid-Open No. 11888/1988).

Problems to be Solved by the Invention However, in the above-mentioned conventional techniques, the main focus is on classifying text areas and non-text areas for character recognition processing, and non-text areas in an image are classified into figures, graphs, etc. ,photograph,
This method has the disadvantage of not being able to classify finely into areas such as line segments and noise areas.

The present invention has been made in view of the above points, and an object of the present invention is to provide an image processing device that automatically extracts and classifies character strings, graphs, photographs, line segments, and noise regions in an image using a simple method. It is an object.

Means for Solving the Problems In order to achieve the above object, the present invention provides character strings 9 from an image.
a rectangular area coordinate extraction unit that extracts rectangular areas of diagrams, photographs, line segments, and noise; a rectangular area feature extraction unit that extracts “features of the rectangular area extracted by the rectangular area coordinate extraction unit”; and the rectangular area feature extraction unit. This image processing device includes a rectangular area classification unit that classifies a rectangular area into character strings, graphs, photographs, line segments, and noise using features extracted from the area.

According to the above configuration, the present invention extracts a rectangular area from an image using the rectangular area coordinate extraction unit, extracts features from the extracted rectangular area using the rectangular area feature extraction unit, and uses the extracted features in the rectangular area classification unit. Character String 9 By comparing the features prepared in advance for each diagram, photograph, line segment, and noise, a rectangular area is transformed into a character string.

Determine whether it corresponds to a diagram, photograph, line segment, or noise.

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 an image processing apparatus according to the present invention. 1 is a character string in the image input section.

Images including charts, photographs, line segments, and noise are scanned and stored in the image memory section 2 as binary signals. 3 is a rectangular area coordinate extraction section with character string 1 chart, photo, a minute.

Extract the coordinates of the smallest rectangular area surrounding the noise.

4 is a rectangular area feature extraction unit, and rectangular area coordinate extraction unit 3
Extract the characteristics of the rectangular area surrounding the extracted character string 2 charts, photographs, line segments, and noise. 5 is a rectangular area classification unit;
The rectangular area feature extraction unit 4 performs extraction and classification.

The image processing apparatus configured as described above will be explained using an input image P shown in FIG. 2 as an example.

The image P input from the image input unit 1 is a character string.

Binary data of 1 for black pixels in graphs, photographs, line segments, and noise areas and 0 for white pixels in background areas is stored in the image memory unit 2.

The rectangular area coordinate extraction unit 3 scans the input image P stored in the image memory unit 2 in the horizontal direction, and if the distance between black pixels is equal to or less than a predetermined threshold value R1, the black pixels are connected to each other. It is assumed that Similarly, when the input image P stored in the image memory unit 2 is scanned in the vertical direction and the distance between black pixels is less than or equal to a predetermined threshold value R2, the black pixels are considered to be connected. . Lateral direction.

Focusing on the connection information between black pixels obtained by scanning in the vertical direction, we calculate the coordinates of the upper left point of the smallest rectangular area (xmi, n yYmi
n) r Lower right point coordinates (xm&! j Ym&
! ). FIG. 3 shows a state in which a rectangular region of a character string is extracted using coordinates. FIG. 4 shows all rectangular areas extracted by the rectangular area coordinate extraction unit 3 from the input image P of FIG. 2.

The rectangular area feature extraction unit 4 calculates the width W of the rectangular area from the rectangular area coordinates of the character strings 9 charts, photographs, line segments, and noise extracted by the rectangular area coordinate extraction unit 3 using equation (1).

W=XmanyXm1n+1 ・・・・・・・・・
(1) Similarly, the height H of the rectangular area is determined from the rectangular area coordinates using the formula @).

H=Y −Y +1 ・・・・・・・・・
...C2) maw min From the width W and height H of the rectangular area, the vertical height V of the rectangular area in the character string direction is determined using equation (3).

From the width W and height H of the rectangular area, calculate the rectangular area size S using the formula (
A).

S=wH・・・・・・・・・・・・←) Width W of rectangular area
and the height H, the rectangular area aspect ratio E is determined by equation (5).

The black pixel density of the rectangular area is calculated from the rectangular area size S and the number B of black pixels in the rectangular area using equation (6).

D = ding (dust) In the rectangular area classification unit 6, - a character string of a numerical document.

For charts, photographs, line segments, and noise, the vertical height V in the character string direction of the rectangular area extracted by the rectangular area feature extraction unit 4 is used.

Classification is performed using the fact that the rectangular area size S, the rectangular area aspect ratio E, and the black pixel density of the rectangular area have specific properties. Specifically, if the vertical height V of a rectangular area in the character string direction is greater than or equal to a predetermined threshold value vthr, the rectangular area is classified as a diagram or a photograph, and if V is not vthr, then the character string, line segment.

It is classified as either noise. A rectangular area classified as a character string, line segment, or noise is determined by a predetermined threshold value of 8t for the rectangular area size S.
If it is hr or more, it is classified as a character string or line segment, and S is 8th
If it is less than r, it is classified as noise. string,
A rectangular area classified as a line segment is classified as a line segment if the rectangular area aspect ratio E is equal to or greater than a predetermined threshold value Ethr, and is classified as a character string if E is not filled with Ethr. A rectangular area classified as a diagram or a photograph is classified as a photograph if the black pixel density of the rectangular area is greater than or equal to a predetermined threshold value Dthr, and is classified as a diagram if D is less than Dthr. FIG. 5 shows an explanatory diagram of the classification conditions for rectangular areas.

The image processing apparatus configured as described above can extract and classify character strings 9, charts, photographs, line segments, and noise from an image containing character strings 2, charts, photographs, line segments, and noise.

Note that by connecting the image processing device of this embodiment to a character recognition device, characters can be extracted and recognized from rectangular areas classified as character strings.

As described in detail, according to the present invention, text strings, graphs, photographs, line segments, and noise regions can be automatically extracted from an input image of a document in an unspecified format using a simple method. . Using this method, character string regions are extracted and recognized character by character using already known character recognition technology, and diagrams, photographs, line segments, and noise regions are each input by performing unique processing. Images can be processed more flexibly, which has great practical effects.

[Brief explanation of drawings]

FIG. 1 is a configuration diagram of an image processing apparatus according to an embodiment of the present invention, FIG. 2 is an explanatory diagram of an input image, FIG. 3 is an explanatory diagram showing rectangular area coordinates of extracted character strings, and FIG. An explanatory diagram showing all rectangular areas extracted for the input image in Figure 2,
FIG. 5 is an explanatory diagram showing classification conditions for rectangular areas. 1... Image input unit, 2... Image memory unit, 3... Rectangular area coordinate extraction unit, 4...
・Rectangular area feature extraction unit, 6... Rectangular area classification unit, 6... Character string area, 7... Line segment area, 8... Photograph area , 9...Chart area,
P...Input image. Name of agent: Patent attorney Shigetaka Awano and 1 other person 1st
Figure / Diagram

Claims (1)

[Claims] an image information input section for inputting image information consisting of at least two sets of elements among character strings, diagrams, photographs, line segments, and noise; and an image information memory for storing the image information input to the image information input section. a rectangular area coordinate extraction unit that extracts a rectangular area of a character string, diagram, photograph, line segment, or noise from the image information stored in the image information memory unit; and a rectangular area extracted by the rectangular area coordinate extraction unit. and a rectangular area classification unit that uses the features extracted by the rectangular area feature extraction unit to classify the rectangular area into character strings, diagrams, photographs, line segments, and noise. An image processing device characterized by: