Daniel E Curtis

US Patent:

8170372, May 1, 2012

Filed:

Mar 30, 2011

Appl. No.:

13/075975

Inventors:

Michael B. Kennedy - Baltimore MD, US
Daniel Curtis - Easton MD, US

International Classification:

G06K 9/36

US Classification:

382291

Abstract:

The present invention is a method and system to precisely locate objects of interest in any given image scene space, which finds the presence of objects based upon pattern matching geometric relationships to a master, known set. The method and system prepares images for feature and attribute detection and identifies the presence of potential objects of interest, then narrows down the objects based upon how well they match a pre designated master template. The method by which matching takes place is done through finding all objects, plotting its area, juxtaposing a sweet spot overlap of its area on master objects, which in turn forms a glyph shape. The glyph shape is recorded, along with all other formed glyphs in an image's scene space and then mapped to form sets using a classifier and finally a pattern matching algorithm. The resulting objects of interest matches are then refined to plot the contour boundaries of the object's grouped elements (arrangement of contiguous pixels of the given object called a Co-Glyph) and finally snapped to its component actual dimensions e. g. , x, y of a character or individual living cell.