Texture image segmentation using combined features from spatial and spectral distribution

Texture discrimination is playing a vital role in a real world image classification and object identification in a content based image retrieval (CBIR) system. For discriminating the textures, exact features have to be extracted. Although there are many techniques available they are not capable of classifying the universal textures because of their inherent limitations. In this paper, a novel method is introduced to extract the features by combining the texture discriminating features of spatial and spectral distribution of image attributes, and a comparison is made with the popular Gaussian and Gabor wavelets based methods for segmenting the image. The segmented outputs and the classification efficiency of the proposed method are found to be better and the time taken is reasonable.