New Similarity Measure for Shape Recognition

• Main Authors: Grouche Lakhdar, Bonton Pierre
• Format: Article
• Language: English
• Published: BİSKA Bilisim Company 2013-03-01
• Series: New Trends in Mathematical Sciences
• Subjects: Distance mask; Histogram equality; Numerical signature; linear approximation; Stationary series; Pattern recognition
• Online Access: http://www.ntmsci.com/ajaxtool/GetArticleByPublishedArticleId?PublishedArticleId=5

Blob shape recognition is used in various application fields such as industrial vision systems, parts recognition, positioning, inspection, etc. This is based on numerical signature which is an effective image processing technique for shape recognition. This paper describe a new fast algorithm for pattern recognition, shapes are being coded by using two vectors, the first one is contour distances and the second one is of the surface histogram, these are working in relative manner. The principle coding is reduced to a product between the mask and the binary image of the shape. The first merit of this algorithm is the high-speed image processing, instead of using images it does operate on vectors. The second merit is the precise recognition of known geometries shapes even for arbitrary or complexes ones.