发明名称 METHOD FOR CHARACTER EXTRACTION FROM COLOR DOCUMENT IMAGE
摘要 PROBLEM TO BE SOLVED: To provide a method for character extraction from color document image that can separate the background color and the character color of an image and extract characters even from a complex and varied background by clustering the color information. SOLUTION: The method for character extraction from color document image, which extracts only character color portions from a color document image with a complex and varied background, removes dithers by smoothing from the color document image, converts the color values of the dither-removed image from the RGB system to the L*u*v* system and prepares a histogram thereof, clusters the color information on a fuzzy basis, prepares color-separated images (binary images) according to the degree of assignment, removes noise from the binary images, provides labels to black pixels and white pixels, then selects a binary image suitable for character extraction, and extracts a line of characters.
申请公布号 JP2002042055(A) 申请公布日期 2002.02.08
申请号 JP20000222063 申请日期 2000.07.24
申请人 JAPAN SCIENCE & TECHNOLOGY CORP 发明人 OKAMOTO MASAYUKI
分类号 G06K9/20;G06T1/00;G06T5/00;G06T7/00;H04N1/40;H04N1/46;H04N1/60 主分类号 G06K9/20
代理机构 代理人
主权项
地址
您可能感兴趣的专利