Method of producing character templates using unsegmented samples

Abstract

A method for producing, or training, a set of character templates uses as the source of training samples an image source of character images, called glyphs, that are not previously segmented or isolated for training. Also used is a labeled glyph position data structure that includes, for each glyph in the image source, a glyph image position in the image source associating an image location of the glyph with a character label paired with the glyph image position that indicates the character in the character set being trained. The labeled glyph position data is used to identify a collection of glyph sample image regions in the image source for each character in the character set; each glyph sample image region is large enough to contain a glyph and typically contains adjacent glyphs for other characters. The invention mathematically characterizes the template construction problem using unsegmented samples as an optimization problem that optimizes a function that represents the set of character templates being trained as an ideal image to be reconstructed to match the input image. The method produces all of the character templates contemporaneously by using a novel pixel scoring technique that implements an approximation of a maximum likelihood criterion subject to a constraint on the templates produced which holds that foreground pixels in adjacently positioned character images have substantially nonoverlapping foreground pixels. The character templates produced may be binary templates or arrays of pixel color probability values, and may also have substantially disjoint supports, such that adjacently imaged templates have substantially no overlapping foreground pixels.