A Novel Feature Descriptor for Image Retrieval by Combining Modified Color Histogram and Diagonally Symmetric Co-occurrence Texture Pattern

In this paper, we have proposed a novel feature descriptors combining color\nand texture information collectively. In our proposed color descriptor\ncomponent, the inter-channel relationship between Hue (H) and Saturation (S)\nchannels in the HSV color space has been explored which was not done earlier.\nWe have quantized the H channel into a number of bins and performed the voting\nwith saturation values and vice versa by following a principle similar to that\nof the HOG descriptor, where orientation of the gradient is quantized into a\ncertain number of bins and voting is done with gradient magnitude. This helps\nus to study the nature of variation of saturation with variation in Hue and\nnature of variation of Hue with the variation in saturation. The texture\ncomponent of our descriptor considers the co-occurrence relationship between\nthe pixels symmetric about both the diagonals of a 3x3 window. Our work is\ninspired from the work done by Dubey et al.[1]. These two components, viz.\ncolor and texture information individually perform better than existing texture\nand color descriptors. Moreover, when concatenated the proposed descriptors\nprovide significant improvement over existing descriptors for content base\ncolor image retrieval. The proposed descriptor has been tested for image\nretrieval on five databases, including texture image databases - MIT VisTex\ndatabase and Salzburg texture database and natural scene databases Corel 1K,\nCorel 5K and Corel 10K. The precision and recall values experimented on these\ndatabases are compared with some state-of-art local patterns. The proposed\nmethod provided satisfactory results from the experiments.\n