Performance comparison of different Wavelet based image fusion techniques for Lumbar Spine Images

Abstract

Image fusion of medical images gives an output representative image that contains more detail than each of the source images, making it an informative medium for clinicians. The main goal of multimodal image fusion is to act as a clinically supportive tool for a better and more accurate diagnosis, so that important information or features are considered. This work is aimed to implement and analyze Discrete Wavelet Transform (DWT) based Image fusion algorithm applied to the Computed Tomography (CT) and Magnetic Resonance Imaging (MRI) images of the lumbar spine. The information available from CT and MRI images totally complements each other, where the former is better for visualization of bony structures and the latter is better for visualization of soft tissues and nerves. Hence, the implemented algorithm effectively uses the information present in CT and MRI images and provides a resultant fuse image that can be used further for diagnosis and treatment planning. The performance analysis of the implemented DWT based image fusion algorithm is evaluated by quantitative quality metrics such as Entropy,  , mutual information, and spatial frequency and also tested under conditions like varying the parameters such as the types of wavelets used for decomposing the input images and the number of decomposition levels. The overall comparison of the majority of metrics has shown that the higher decomposition value in the wavelet of each family performs better in all of the cases presented in the study.

URN:NBN:sciencein.jist.2024.v12.703