New design of adaptive Gabor wavelet filter bank for medical image retrieval

Gabor wavelet is widely used in the analysis of texture feature. This study presents a novel feature descriptor based on the design of adaptive Gabor wavelet filter-bank for medical image retrieval. The design of proposed Gabor wavelet provides flexibility to extract the dominant directional features from medical images. First, peaks in the spectrum of medical image are analysed to determine the dominant directions present in the image. With these dominant directions, a bank of Gabor-filters is designed to extract the directional features effectively. Next, feature vector is derived by computing the energy and standard deviation from the Gabor filtered coefficients at a particular scale and orientation. The use of maximum likelihood estimation (MLE) is suggested to measure the similarity between the feature vectors of heterogeneous medical images. The performance of the proposed method is evaluated on three different publicly available databases namely NEMA, OASIS and EXACT09. The performance in terms of average retrieval precision (ARP), average retrieval rate (ARR) and computational time are compared with well-known existing methods. It is observed from experimental results that the proposed approach achieved ARP of 85.32, 85.24 and 77% and ARR of 31.33, 14.05 and 23.78% for NEMA, OASIS and EXACT09 databases respectively.