EP0516576A2
Expired / Lapsed
EP
⏱️ 0.0 years remaining
Method of discriminating between text and graphics
Assignee: Scitex Corp Ltd
•
Filing: 1992-05-21
•
Granted: —
•
Expires: 2012-05-21
Abstract
Abstract A method of analyzing a source image to separate text from graphics, by: (a) scanning and digitizing the source image to obtain a binary image including black and white objects; (b) filtering out the noise from the binary image; (c) extracting the contours therefrom of the black objects and the white objects; (d) evaluating inclusion relationships between the objects, and generating a tree-like structure of such relationships; (e) utilizing the contours for measuring the objects to obtain the shape properties of each object; (f) effecting classification of the objects as graphics or text according to the measured shape properties and then generating tree-like structure of the inclusion relationships; (g) and utilizing the source image and the classification of the objects for generating outputs representing graphics and text, respectively.
Inventors
- 👤 Assaf Zeira
Bibliographic Details
Application Number
—
Jurisdiction
EP
Data Source
google_patents
Last Synchronized
2026-11-08 14:41
CPC Classifications
G; G06; G06T; G06T7/00; G06T7/10; G06T7/12; G06T7/194; G06V; G06V30/00; G06V30/40; G06V30/41; G06V30/413; G06T2207/00; G06T2207/10; G06T2207/10004; G06T2207/10008; G06T2207/30; G06T2207/30176
Claims Analysis
Total Claims: 15
Claim 1
Independent Claim
138 words
A method of analyzing a source image to separate text from graphics, comprising: (a) scanning and digitizing the source image to obtain a binary image including black and white objects; (b) filtering out the noise from the binary image to obtain a filtered binary image; (c) extracting the contours of the black objects and the white objects from the filtered binary image; (d) evaluating inclusion relationships between the objects, and generating a tree-like structure of such relationships; (e) utilizing said contours for measuring the objects to obtain the shape properties of each object; (f) effecting classification of the objects as graphics or text according to the measured shape properties and then generating tree-like structure of the inclusion relationships; (g) and utilizing said source image and said classification of the objects for generating outputs representing graphics and text, respectively.
Claim 2
Dependent Claim
21 words
The method according to Claim 1, wherein in step (b), the noise is filtered out by dilation of the black pixels.
Claim 3
Independent Claim
32 words
The method according to either of Claims 1 or 2, wherein in step (e), the objects are measured in a top-down sequence, starting with the object at the root of the tree.
Claim 4
Independent Claim
50 words
The method according to any one of Claims 1-3, wherein in step (c), extracting the contour of the black objects and the white objects from the filtered binary image is effected by a single scan in which a window is convolved with the filtered binary image in a raster fashion.
Claim 5
Dependent Claim
65 words
The method according to Claim 4, wherein the window scans the image along a line and returns an indication of the type of pattern seen from the window and an indication of the center of the window, each type pattern being processed differently to determine whether a new object is started, continued or ended, all objects intersecting the current scan line being processed in parallel.
Claim 6
Dependent Claim
54 words
The method according to Claim 5, wherein a maximal point encountered during the window scan is considered to be a starting point of a new object, but if later the scan indicates it was a maximal point of a previously indicated object, the new object is merged with that of the previously indicated object.
Claim 7
Independent Claim
62 words
The method according to any one of Claims 1-6, wherein in step (d), the tree-like structure is generated by setting a pointer from each object to its predecessor, the predecessor of an object being found by determining which of the object contours is the outermost one, and then setting the pointer to point at the object on one side of that contour.
Claim 8
Independent Claim
39 words
The method according to any one of Claims 1-7, wherein in step (e), the objects are measured to obtain the following shape properties of each object: area of the object, number of contours, and perimeter length of each contour.
Claim 9
Dependent Claim
35 words
The method according to Claim 8, wherein in step (e), the following additional properties are determined from the measured shape properties: elongation, hull area, hull eccentricity, black/white ratio, Euler number, and number of sharp corners.
Claim 10
Dependent Claim
68 words
The method according to Claim 9, wherein the number of sharp corners is determined by: generating several polygonal approximations of the contour, with each generation having a bigger error threshold, as long as the number of polygon segments drops linearly with respect to the increase in the error threshold; and determining that a sharp corner exists when the last polygon approximation has an angle of less than 60°.
Claim 11
Independent Claim
27 words
The method according to any one of Claims 1-10, wherein in step (g), the generated outputs represeting graphics and texts are in the form of different images.
Claim 12
Independent Claim
31 words
The method according to any one of Claims 1-10, wherein in step (g), the generated outputs representing graphics and texts are in the form of different colours of the same image.
Claim 13
Independent Claim
20 words
The method according to any one of Claims 1-10, wherein the source image contains text of different sizes, orientationsand/or fonts.
Claim 14
Dependent Claim
47 words
The method according to Claim 2, wherein steps (a)-(g) are repeated, except the noise is filtered out in step (b) by dilation of the white pixels, so that white objects of the source image are separated, thereby providing discrimination of white text and graphics over black background.
Claim 15
Independent Claim
31 words
The method according to any one of Claims 1-10, wherein the source image contains black text, white text, black graphics, white graphics, black and white background, and black and white noise.
References Cited (Backward Citations)
No citation records indexed.
USPTO Annuity & Maintenance Windows
Maintenance schedule is calculated for granted US utility patents.