Content Based Image Retrieval Using Color , Texture And Hybrid Features 33

Image Retrieval system is an effective and efficient tool for managing large image databases. In content based image retrieval system, user provides a query image in order to retrieve relevant images stored in the database. Feature extraction is an important step for extracting valuable information from the image. Color is one of the most important low-level features used in most of the content based image retrieval systems. However, image retrieval using only color features often provide very unsatisfactory results because in many cases, images with similar colors do not have similar content. As a solution for this problem, proposed approach describes a novel algorithm for content based image retrieval based on color and texture features. The proposed algorithm also generates a feature vector that combines both color and texture features. For extracting color features, histogram and color moments are used in order to represent global color distribution and to include spatial color information, color autocorrelogram is used. Wavelet decomposition is used to reduce size of the feature vector and simultaneously preserving the content details for texture feature extraction.To take advantage of their strong orientation selectivity, log Gabor filter is also adopted to extract texture feature. The robustness of the system is tested against query image. Wang’s image database is used for experimental analysis and results are shown in terms of precision.