A Deep Optimization Image Deblurring Model Based on Internal and External Information
Yongqun Tan, Lingli Zhang · 2024
In recent decades, there has been a notable focus on addressing image deblurring problem. Different deblurring algorithms integrate prior information into the image deblurring model, and the effectiveness of the prior information directly determines the performance of the deblurring. This paper presents a deep optimization image deblurring model based on internal and external information, which is joint with total variation, non-local self-similarity and convolutional neural network. The corresponding algorithm of the presented model utilizes half quadratic splitting method to decouple the fidelity and regularization terms, which can be solved through the sub-problems, respectively. Numerical experiments show that the presented algorithm can outperform the existing algorithms both in preserving the edges and assessment indicators (PSNR, RMSE, and SSIM).