Medical Image Fusion Based On Modified Gradient Weighted Sum And FPGA Implementation Research

Multimodal medical image fusion refers to the combination of multiple images from a single imaging modality or from multiple imaging modalities.The significance of fusion is that it can effectively merge the complementary information of multimodal medical images,thus avoiding the drawbacks of limited information in unimodal medical images.Fusion of images can improve diagnostic accuracy so that physicians can make more fair and objective decisions.Therefore,multimodal medical image fusion technology plays an increasingly influential role in clinical diagnosis.This paper focuses on the fusion of multimodal medical images,and the main research contents are as follows:(1)To address the problems of insufficient texture and edge detail extraction and low contrast of existing medical image fusion algorithms,this paper proposes a multimodal medical image fusion algorithm based on modified gradient weighted sum.The algorithm performs multiscale decomposition using Laplace pyramid,and the decomposed high-frequency components are fused using the proposed modified gradient weighted sum algorithm,while the lowfrequency components are fused using the sparse representation algorithm.After obtaining the fused high and low frequency components,the final fused image can be obtained by reconstructing them using the Laplace pyramid.(2)Multimodal medical image fusion can be divided into two cases,such as the fusion of grayscale images with grayscale images for CT-MRI and the fusion of grayscale images with color images for MRI-PET and MRI-SPECT.To effectively avoid color distortion,two different fusion frameworks are designed based on the proposed fusion algorithm.Experiments show that the proposed fusion algorithm achieves excellent performance in both subjective and objective evaluations of fusion results and has high fusion efficiency compared with Laplace pyramid fusion,sparse representation fusion and the latest research.(3)To address the problem that the fusion efficiency of medical image fusion algorithm is low and insufficient to meet the demand of real-time applications,the proposed multimodal medical image fusion algorithm is further implemented in FPGA.In order to speed up the processing speed of the algorithm,the designed hardware architecture is implemented in the form of pipeline.The designed hardware architecture based on the modified gradient weighted sum fusion algorithm mainly contains a pyramid decomposition module,a fusion module,and a pyramid reconstruction module.The experiments show that the designed hardware architecture can achieve similar fusion effect as PC side and can significantly improve the fusion efficiency of the algorithm to meet the demand of real-time applications.