An experimental comparison of features in content based image retrieval system

Abstract

An investigation comparative of many descriptors of various images in content-based image retrieval system (CBIR) is described in the paper. This paper describes more number of various features in CBIR system and compare the four different Color and texture based existing low level Feature Extraction Techniques such as Tamura Texture Features, RGP Color Histogram, Gabor Features and Joint Picture Editor Group (JPEG) Coefficients Histogram. The Proposed Techniques such as Fuzzy color and texture histogram (FCTH) and Color and edge directivity descriptor (CEDD) which retrieve the relevant images matching with edge, texture and color value from the Corel library. The Haar wavelet transform (HWT), Discrete wavelet transform (DWT) and algorithm of Fuzzy linking with Gabor filter are used in the proposed paper. The proposed approach gives good result in Average Image Retrieval precision (IRP) and Recall value.