Integrated CBIR Using Texture, Fourier Descriptor and Color Histogram

Abstract

Content-Based Image Retrieval (CBIR) considers the
characteristics of the image itself, for example its shapes, colors and textures. The Current approaches to CBIR differ in terms of which image features are extracted. Recent work deals with combination of
distances or scores from different and independent representations. CBIR has many application fields such as, education, commerce, military, searching, biomedicine, and web image classification. This
paper proposes a new image retrieval system, which uses color and Shape descriptions information and Texture to form the feature vectors. This framework integrates the ycbcr color histogram which represents the color feature, Fourier descriptor as shape descriptor
and Edge Histogram as texture descriptor to enhance the retrieval results. The proposed technique is proper for precisely retrieving images even in deformation cases such as geometric deformations
and noise. It is tested on a standard image databases such as Wang and UCID database. Experimental work show that the proposed approach improves the precision and recall of retrieval results
compared to other approaches reported in literature.