Text/image separation method

Claim 2

A parallel-architecture method for separation of text segment and image segment contained in a document, said method comprising the following steps: (1) scanning said document by gray scale scan-in means as an input gray scale image, said input gray scale image comprising a plurality of pixels, each having a gray level belonging to a set of successive gray levels; (2) halftone-processing said input gray scale image to convert said gray level pixels into black/white pixels, to create the illusion of gray by means of an appropriate arrangement of black and white dots, and to produce a first signal representing the halftone processing result of a currently-processed gray level pixel; (3) constructing a window which covers at least the currently-processed pixel and a plurality of pixels surrounding the currently-processed pixel; (4) detecting whether edges exist between the currently-processed pixel and the surrounding pixels by means of reading said gray-level pixels from said gray-scale scan-in means, subtracting in sequence the gray-level of the currently-processed pixel with those of said surrounding pixels to obtain absolute differences thereof, comparing said absolute differences with a pre-determined edge threshold to determine the existence of edges, generating a second signal to indicate the detection, providing a surrounding pixel priority to select one of said surrounding pixels when a plurality of edges exist, calculating a cutting threshold by taking average of the gray levels of the currently-processed pixel and the selected surrounding pixel, and producing a third signal, which represents an output state of the currently-processed pixel, by comparing the gray level of the currently-processed pixel with said cutting threshold; (5) selecting said first signal or said third signal in accordance with said second signal; and (6) printing out in binary digit said selected signal.

Claim 3

The parallel-architecture method as claimed in claim 1, wherein the size of said window is three pixels by three pixels.

Claim 4

The parallel-architecture method as claimed in claim 1, wherein an edge is described when the absolute difference of gray level between the currently-processed pixel and one of said surrounding pixels is greater than said edge threshold.

Claim 5

The parallel-architecture method as claimed in claim 1, wherein said edge threshold is determined by scanning a piece of white paper with black strips intervening thereon, obtaining the gray level variations on transition areas between said black strips and the white background to establish a statistical distribution thereof, and selecting at least value within three standard deviations of the distribution as said edge threshold.

Claim 6

The parallel-architecture method as claimed in claim 1, wherein said surrounding pixel priority is determined according to the sequences from first to eighth: the pixel on top-left position, that on top-middle position, that on top-right position, that on middle-left position, that on middle-right position, that on bottom-left position, that on bottom-middle position and that on bottom -right position of said window.

Claim 7

A parallel-architecture system for separation of text segment and image segment contained in a document, said system comprising: (1) a gray-scale scan-in means for scanning said document as an input gray scale image, said input gray scale image comprising a plurality of pixels, each having a gray level belonging to a set of successive gray levels; (2) a halftone processing means for converting said gray level pixels of said document after scanned by the gray-scale scan-in means into black/white pixels, creating the illusion of gray by means of an appropriate arrangement of black and white dots, and producing a first signal which is represented the halftone processing result of a currently-processed pixel; (3) a window which covers at least the currently-processed pixel at a center position therein and a plurality of pixels surrounding said currently-processed pixel; (4) an edge detection means for reading said gray-level pixels, subtracting in sequence the gray-level of the currently-processed pixel centered in the window with those of said surrounding pixels to obtain the absolute differences thereof, comparing said absolute differences with an edge threshold to decide the existence of edges, generating a second signal to indicate the detection of the edges, providing a surrounding pixel priority to select one of the said surrounding pixels when a plurality of edges exist, calculating a cutting threshold by taking average of the gray levels of the currently-processed pixel and the selected surrounding pixel, and producing a third signal, which represents an output state of the currently-processed pixel, by comparing the gray level of the currently-processed pixel with said cutting threshold; (5) a selector means for selecting said first signal from said halftone processing means or said third signal from said edge detection means in accordance with said second signal; and (6) a binary print-out means for printing out said selected signal by said selector means.

Claim 8

A parallel-architecture system for separation of text segment and image segment contained in a document, said system comprising: (1) a gray-scale scan-in means for scanning said document as an input gray scale image, said input gray scale image comprising a plurality of pixels, each having a gray level belonging to a set of successive gray levels; (2) a halftone processing means for converting said gray level pixels of said document after scanned by the gray-scale scan-in means into black/white pixels, creating the illusion of gray by means of an appropriate arrangement of black and white dots, and producing a first signal which is represented the halftone processing result of a currently-processed pixel; (3) a window which covers at least the currently-processed pixel at a center position therein and a plurality of pixels surrounding said currently-processed pixel; (4) an edge detection means for reading said gray-level pixels, subtracting in sequence the gray-level of the currently-processed pixel centered in the window with those of said surrounding pixels to obtain the absolute differences thereof, comparing said absolute differences with an edge threshold to decide the existence of edges, generating a second signal to indicate the detection of the edges, providing a surrounding pixel priority to select one of the said surrounding pixels when a plurality of edges exist, calculating a cutting threshold by taking average of the gray levels of the currently-processed pixel and the selected surrounding pixel, and producing a third signal, which represents an output state of the currently-processed pixel, by comparing the gray level of the currently-processed pixel with said cutting threshold; (5) a selector means for selecting said first signal from said halftone processing means or said third signal from said edge detection means in accordance with said second signal; and (6) a binary print-out means for printing out said selected signal by said selector means.

Claim 9

The parallel-architecture system as claimed in claim 6, wherein said window is a 3 pixels by 3 pixels window, and said edge detection means comprises: (1) a storing means for storing the gray levels of the pixels contained in said window; (2) a subtractor means for subtracting in sequence the gray level of the currently-processed pixel centered in said window with those of said surrounding pixels, and for calculating the absolute differences so obtained; (3) an edge threshold comparator means for comparing said absolute differences respectively with a pre-set edge threshold to detect if edges exist between the currently-processed pixel and the surrounding pixels and generating the second signal to indicate the detection of the edges, said edge threshold comparator means further providing a surrounding pixel priority; (4) a priority selection means for selecting one of the surrounding pixels that have edges existing between them and the currently-processed pixel in accordance with the surrounding pixel priority; (5) a calculation means for calculating the cutting threshold by taking average of the gray levels of the currently-processed pixel and the selected surrounding pixels; and (6) a comparator means for comparing the gray level of the currently-processed pixel with the cutting threshold to provide the third signal.

Claim 10

The parallel-architecture system as claimed in claim 6, wherein said edge threshold is determined by scanning a piece of white paper with black strips intervening thereon, obtaining gray level variations on transition areas between said black strips and the white background to establish a statistical distribution thereof, and selecting a least value within three standard deviations of the distribution as said edge threshold.

Claim 11

The parallel-architecture system as claimed in claim 6, wherein said edge detection means decides said edges exist between the currently-processed pixel and the surrounding pixels if the corresponding absolute differences are greater than said edge threshold.

Claim 12

The parallel-architecture system as claimed in claim 6, wherein said surrounding pixel priority is determined according to the sequences from first to eighth: the pixel on top-left position, that on top-middle position, that on top-right position, that on middle-left position, that on middle-right position, that on bottom-left position, that on bottom-middle position and that on bottom-right position of said window.