Image indexing and retrieval using Gabor wavelet and Legendre moments

As the volume of images grows at an amazing speed, efficient management of image databases becomes more and more important. A practical way to index an image database is to use low-level image features such as textures or colors which can be attracted by machines automatically. This paper presents a new method of image indexing and retrieval based on Gabor wavelet and Legendre moments called (GWLM). It is well known that Gabor wavelet decomposition achieves the theoretical lower bound of the uncertainty principle. They attain maximum joint space-frequency resolution which is highly significant in the process of texture extraction in which the conflicting objectives of accuracy in texture representation and texture spatial localization are both important. To improve the efficiency of retrieval rate we also utilized the Legendre moments which significantly improved the retrieval rate up to about 25%. The feature vector consists, of 48 parameters from Gabor wavelets plus 10 Legendre moments. The length of feature vector is relatively small compared to other methods such as WBIIS which uses 768 features. This has a significant impact on the speed of retrieval process. Experimental results carried out on two databases of natural and medical images. The proposed method clearly outperforms some image indexing methods such as WBIIS and DHSE. The presented method works independent of image color or brightness, therefore it is able to retrieve similar images with different colors or brightness.