Structured Local Binary Haar pattern for graphics retrieval

Feature extraction is an important issue in graphics retrieval. Local feature based descriptors are currently the predominate method used in image retrieval and object recognition. Inspired by the success of Haar feature and Local Binary Pattern (LBP), a novel feature named structured local binary Haar pattern (SLBHP) is proposed for graphics retrieval in this paper. SLBHP encodes the polarity instead of the magnitude of the difference between accumulated gray values of adjacent rectangles. Experimental results on graphics retrieval show that the discriminative power of SLBHP is better than those of using edge points (EP), Haar feature, and LBP even in noisy condition.