Content Based Image Retrieval

Abstract—Content Based Image Retrieval. We have proposed a novel representation to describe color, intensity, edge orientation, frequency and spatial layout as histogram-based features via simulating human’s visual mechanism. In the representation, Color volume is used as a low-level feature to detect salient regions. At the same time, a novel representation method of visual feature, namely Cauchy density function histogram, is used to describe color image content. It has the ability of integrating low-level features, frequency and spatial information. The experimental results of CBIR have indicated that Cauchy density function histogram outperform Bow histogram and multi-texton histogram. We have proposed a novel representation to describe color, < Final Year Projects > intensity, edge orientation, frequency and spatial layout as histogram-based features via simulating human’s visual mechanism. In the representation, Color volume is used as a low-level feature to detect salient regions. At the same time, a novel representation method of visual feature, namely Cauchy density function histogram, is used to describe color image content.