Statistical framework for content-based medical image retrieval based on wavelet orthogonal polynomial model with multiresolution structure

Abstract

This paper proposes wavelet based orthogonal polynomial coefficients model for content based image retrieval (CBIR). The coefficients are categorized into low-frequency and high-frequency based on a criteria. The criteria is adaptively determined and fixed according to the nature and structure of the image, because the wavelet based orthogonal polynomial model spatially localizes the frequency information. Wavelet packet and Daubechies-4 transforms are jointly used to construct both approximation (low-frequency) and detailed (high-frequency) multiresolution image subbands. Color feature are extracted from low-frequency subband based on color autocorrelogram, whereas texture features are extracted from high-frequency subband based on co-occurrence matrix. Based on these features, the feature vector is formed. The proposed CBIR method reduces the feature variation when different modalities of images are combined. The proposed system assessed two medical image databases and one general image database with Minkowski-form distance method. The experimental results show that the proposed method achieves comparable retrieval performance with medical dataset; moreover, it is very fast with low computational load. Further, the obtained results were compared with other recently developed methods such as highly adaptive wavelet method, Wavelet optimization method and effective CBIR techniques. The proposed method yields better results compared to that of existing methods.