Design of Dyadic Wavelet Filter Bank Through Lifting Scheme for Medical Image Retrieval

Medical image retrieval is an important component in enabling clinical decisions by quickly and accurately accessing similar cases. It can be optimized, and we introduce a dyadic wavelet filter bank based on the lifting scheme—a scheme that is particularly efficient and flexible for the design of the wavelet transforms. Our method exploits dyadic scaling for multilevel frequency decimation, which enhances significant image features such as texture and edges. Such features are particularly significant for medical imaging, where small variation can be symptomatic. In comparison to fixed wavelet filtering, the lifting design lets the wavelet coefficients be tailored, and thus it is more sensitive to varying contrasts and noise over different sets of clinical data. We evaluated the suggested algorithm using benchmark clinical image collections and found it to bring consistent improvement over retrieval performance. Significant retrieval metrics such as average retrieval precision (ARP) and average retrieval rate (ARR) outclassed conventional techniques based on wavelets. The lifting scheme also cut down on computational overhead, which resulted in the method being applicable for real-time or large-scale retrieval systems. This work, overall, serves as a demonstration of how an optimized wavelet design can considerably advance retrieval for medical images, leading the way for more effective and responsive image-based diagnostic solutions.