Content Based Image Retrieval Using 3D Center Symmetric Local Binary Co-occurrence Pattern

Abstract

This paper presents a novel algorithm three dimensional center symmetric local binary co-occurrence pattern (3DC-SLBCoP) for retrieval of images. Standard local binary pattern and its forms uses 2D plane of the image. On the other hand, proposed method leads to 3D volume by extracting Gaussian filtered images using multiresolution Gaussian filter banks and computes the relationship between center pixel and its neighbors in five selected directions. Center symmetric local binary pattern (CSLBP) image is formed by encoding the relationship between focus pixel and its center symmetric neighboring pixels. Thus, gray level co-occurrence matrix (GLCM) of the CSLBP map in four directions leads to the formation of feature vector. Experiments are performed and results are analyzed on benchmark datasets. Analyzed retrieval results clearly better than the other well known methods by considering average retrieval precision and average retrieval rate as evaluation measures.