Local co-occurrence pattern for color and texture image retrieval

摘要

Local patterns have two problems: 1) the traditional local patterns methods only consider the frequency of each pattern, and does not consider the co-occurrence information between adjacent pixels pairs in the image; 2)the traditional methods limit on the gray texture analysis, ignoring the importance of color information. To address above problems, a novel method is proposed for color image retrieval using CLGBCOP (Color Local Gabor binary Co-occurrence Pattern). In this work, global color information is extracted by using color quantization, and the combination of LBCOP (Local binary Co-occurrence Pattern) and Gabor Transform is applied to extract texture information. LBCOP encodes the co-occurrence of local binary edges between center pixel and its neighbors along different directions. Extensive experiments are conducted on Corel-1000, Corel-5000 and Corel-10000 databases to verify the feasibility of the proposed method.