Brain Image Fusion Approach based on Side Window Filtering

Abstract

Brain medical image fusion plays an important role in framing a contemporary image to enhance the reciprocal and repetitive information for diagnosis purposes. A novel approach using kernel-based image filtering on brain images is presented. Firstly, the Bilateral filter is used to generate a high-frequency component of a source image. Secondly, an intensity component is estimated for the first image. Thirdly, side window filtering is employed on several filters, including the guided filter, gradient guided filter, and weighted guided filter. Thereby minimizing the difference between the intensity component of the first image and the low pass filter of the second image. Finally, the fusion result is evaluated based on three evaluation indexes, including standard deviation (STD), features mutual information (FMI), average gradient (AG). The fused image based on this algorithm contains more information, more details, and clearer edges for better diagnosis. Thus, our fused image-based method is good at finding the position and state of the target volume, which leads to keeping away from the healthy parts and ensuring patients’ soundness.