Optimized rotation invariant content based image retrieval with local binary pattern

Growth of the image mining arena calls for the need of quality image retrieval techniques in par with the human perception which are invariant to scale and rotation. An optimized content based image retrieval system based on local visual attention features to bridge the semantic gap problem is proposed. The approach involves the salient point detection using Scale Up Robust Features (SURF) detector. Feature vector characterizing the interest points immune to rotation include the extraction of correlogram as color feature, a new texture pattern named Optimized Rotational invariant Local Binary Pattern (OR-LBP) with high dimensionality reduction as texture feature and the area of convex hull as shape feature. Similarity matching technique is implemented with minimum Manhattan distance between query image and database image. Experimental results in this paper demonstrate the optimized performance of the proposed approach with consistent precision.