Abstract
In this paper, we describe a medical image fusion technique based on relative total variation decomposition (RTVD) that can concurrently maintain the texture and contrast information of the input images. The source images are initially separated into structural and texture components based on the relative total variation. The former is mostly composed of the large frame structure and brightness of the source images, while the latter is composed of the texture and noise with low gradient values. Second, distinct fusion weights are created utilizing traits of the structure and texture layers. To maintain the texture information, the weights of texture parts are calculated in accordance with the saliency map, and the weights of structure parts are established in accordance with image energy to maintain the brightness of original images. The fused image could eventually be recreated utilizing sub-images and weights that were previously gathered. We also conduct qualitative and quantitative experiments to verify the effectiveness of the RTVD approach utilizing publically accessible datasets. The findings demonstrate that the RTVD fusion technique performs better than various more sophisticated algorithms in terms of maintaining contrast, preventing edge blurring, and lowering noise. The fusion outcome also more closely matches disease diagnosis.