基于选通图像的多焦点图像融合算法研究OA

Gated imaging,leveraging the layer-by-layer scanning and imaging characteristics of gated cameras for scenes at different depths,exhibits significant advantages in target recognition and structure restoration in complex scenarios.However,constrained by system noise,single-exposure conditions,and hardware resolution,gated images generally suffer from issues such as low signal-to-noise ratio(SNR),local blurriness,and incomplete detailed information,which severely restrict image quality and the accuracy of subsequent information extraction.Multi-focus fusion technology can effectively alleviate image blurriness and structural defects,yet existing methods still show limitations in focus representation,boundary modeling,and semantic interaction for grayscale images.To address these problems,we propose a multi-focus fusion network for gated images,named MGIF-Net(Multi-focus Gated Images Fusion Network).Firstly,a lightweight SE-ResNet-based shallow attention module is adopted to enhance the expressive ability of focus regions and suppress redundant information.Secondly,the Deep-Fuse deep feature interaction module is introduced to strengthen the fusion of boundary structures and semantic information.Finally,a multi-scale feature fusion module(MSFF)is constructed to realize cross-scale joint modeling,thereby improving structural consistency and detail preservation.Experiments conducted on gated image datasets,Lytro,and MFFW datasets demonstrate that MGIF-Net outperforms existing methods in objective evaluation metrics such as SSIM and information entropy.It also exhibits stronger robustness in blurred boundary restoration and detail recovery in complex background regions,which verifies the effectiveness and advancement of the proposed model.This research provides reliable technical support for en-hancing the application value of gated images.