Skip to Main Content
Content-Based Image Retrieval (CBIR) allows automatic extraction of target images from a database of images according to objective visual contents of the image itself. Texture analysis is a popular operation for CBIR. In this paper we propose a texture analysis based scheme for Color Image Retrieval. We use Haar wavelet to decompose color images into multilevel scale and wavelet coefficients. From the data of low frequency channel we extract features by means of F-norm theory and from data of high frequency channels we extract statistical features by computing gradient directions. Feature vectors extracted from images are then subjected to fuzzy logic based similarity comparison to obtain degree of similarity. A database of 100 color images each of size 256 ?? 256 pixels is used for evaluation. The results are encouraging with best retrieval accuracy of 90.33%.