Image Retrieval Using Local Object Appearance and Spatial Color Layout

In the current scenario, information retrieval is one of the major research areas. The retrieval process in image databases, evaluation of similarity and their mutual relationships depends on the content representation with various features. Since 1992, the multidimensional research has been carried out to meet out the aim with color, shape, texture and spatial layout as the features. In this paper, we provide a spatial and spectral layout of the images with CPAM (Colored Pattern Appearance Model) and a descriptor for characterizing local object appearance and shape using Histogram of Oriented Gradients (HOG). In particular, to get rid of the errors in images due to occlusions, facade and lighting changes, we propose to extract HOG descriptors from the images in regular intervals. Second, fusion of HOG descriptors at different scales allows capturing important structure for image retrieval. Third, the CPAM helps to fully exploit the human vision system. Analytical comparison and experimental results show that the proposed look ahead improves the state-of-the-art in state space search methods. Keywords— Content Based Image Retrieval, HOG, CPAM.