Edge and center symmetric dual cross pattern based shape recognition model

Abstract

The pixels are generally divided into two classes and these are textural pixels and edge pixels. Directional vectors are used to describe the edge pixels and local patterns are used to describe textural pixels. In this article, a new shape recognition model with edge and center symmetric dual cross pattern (CSDCP) is proposed for shape definition. The proposed model consists of edge extraction, edge direction coding, CSDCP, histogram extraction and classification stages. Canny and Laplacian of Gaussian (LoG) edge extraction operators are used in edge extraction. The extracted edges are encoded using direction encoding. The obtained directions were applied with one dimensional CSDCP and 8 bit values were obtained. The histograms of these values are calculated and the feature vector is obtained. The K nearest neighbors, linear support vector machine and linear discriminant analysis were used to classify features. The Kimia99 and MPEG7 databases were used to test the performance of the proposed model. The results show that the proposed model is a successful model for shape recognition.