Image defogging algorithm based on multi-scale residual feature fusion
Aiming at the problems of dim color,poor visual fidelity and loss of detail features of the image after processing by existing defogging algorithms,this paper proposes an image defogging algorithm based on multi-scale residual feature fusion.Firstly,a multi-scale parallel feature layer is designed to extract image features from different scales to improve the robustness of the network.Then,the residual network connection layer is designed to realize the transmission and connection of information between multiple convolutional layers,improve the feature utilization rate and speed up feature extraction.The depth feature information fusion layer embedded in the attention mechanism is designed to focus on the key information of the image.It can effectively improve the clarity of the image and reduce the background noise interference.Finally,a color correction and enhancement method based on fog removal theory and exposure fusion is designed to solve the problem of dim image color after network defogging.The experimental results show that the proposed defogging enhancement algorithm achieves the peak signal-to-noise ratio(PSNR),structural similarity(SSIM)and mean square error(MSE)of 21.37 dB,82%and 473.6 on the public data sets SOTS,OTS and RTTS,respectively,which effectively improves the image quality degradation caused by foggy weather with better performance.