Multi-feature histogram intersection for Efficient Content Based Image Retrieval

This paper presents a multi-feature technique that integrates three different types of histograms for Efficient Content Based Image Retrieval. Color feature is extracted in the form of color histogram using Hue, Saturation and Value (HSV) color space. Local primitives of texture are extracted using Local Binary Pattern. The shape information is obtained using edge histograms computed for three different edge orientations. Histograms are then compared using three different distance measures and the results are tabulated. Finally a weighted similarity index is computed for refining the retrieval. The results depict the superiority of the proposed algorithm over the techniques using various other measures to represent color, texture and shape. The method provides an average retrieval accuracy of 70% on Wang´s Image Database.