Image Retrieval Based on Block Motif Co-Occurrence Matrix

Motif co-occurrence matrix (MCM) is one of the commonly used image features descriptions. However, MCM has two shortcomings. One is that it doesn’t meets translation invariance, and another is that different sub-blocks can be represented by the same motif. In order to overcome the two shortcomings, an image retrieval method based on blocked motif co-occurrence matrix (BMCM) is proposed. BMCM divides the image into five regions firstly, and then it extracts the quantized HSV color histogram feature, MCM feature and local binary pattern feature from each of the five regions. Considering that different attributes and contents of the image are described by different characteristics, this paper achieves image retrieval through a weighted fusion of the above three features. Experimental results in Corel 1k standard image library show that the proposed method has higher precision and lower computation complexity compared with MCM, BCTF and MCMCM algorithm.