Lifting scheme based image retrieval in medical CT and MRI databases

In content based retrieval system, the reliability of image retrieval outcomes relies much upon the image features utilised for measurement of image similarity. In this work, we present retrieval of medical images based on lifting scheme wavelet transform. Low computational complexity in terms of integer-to-integer transform is the main advantage of lifting scheme over classical wavelet transforms which are implemented using floating point operation. In this work, the features are extracted from the medical images using lifting scheme. The feature vector, used to measure similarity, is determined by evaluating the standard deviation and energy from the lifting based wavelet filtered coefficients. The retrieved image is the pertinence between the query image and the image from database, the ranking of the similarity relevance is as per the closest similarity determined by Manhattan distance. The proposed work performance is assessed on three different publicly available medical image databases namely NEMA, OASIS and EXACT09. The effectiveness of the proposed method is compared with other well known existing methods in terms of average retrieval precision (ARP) and average retrieval rate (ARR).