基于加权TV/SAR联合先验与最小线性KL散度的图像重构算法

为了消除当前图像重构算法存在的振铃效应,避免过度平滑图像纹理区域,使其兼顾较好的细小边缘保持与丰富纹理,以获取较高的重构图像视觉质量,提出了基于加权TV(total variation)/SAR(simutanneous auto-regression)联合先验与最小线性KL散度凸组合的图像重构算法.引入权重因子,从退化图像中提取出非局部SSIM约束,联合TV函数,设计加权TV图像正则先验,增强稀疏性;根据SAR先验与加权TV正则先验,获取重构图像的联合后验分布;再建立最小线性KL散度函数凸组合,并引入最优最小化技术,求解后验分布,完成贝叶斯推理.并研究了本文算法在不同退化程度下的用户响应.测试结果显示:与当前图像重构技术相比,本文算法的复原效果较为理想;在图像受损严重时,本文算法更受用户欢迎.

Image Reconstruction Algorithm Based on Combining Priori Model of Weighted TV-SAR and Minimizing the Linear Convex Combinations of Kullback-Leibler

In order to eliminate ringing effect existed in current image reconstruction algorithm and avoid over-smooth in image texture area for maintaining fine edge and rich texture to obtain the high reconstruction image visual quality,the image reconstruction algorithm based on weighted TV/SAR combined priori and minimizing linear convex combination of KL divergence functions is proposed.The weighted TV image regular prior is designed by introducing weighted factor and extracting the non-local constraints from degraded image,as well as combination with the TV function to enhance sparsity.Then the posterior distribution of the reconstructed image is generated by combination with SAR priori and weighted TV regular prior.Finally,the posterior distribution is solved to complete Bayesian inference by building minimizing linear convex combination of KL divergence functions and introducing the majorization-minimization(MM) technology.Additionally,the user response of this algorithm under different degrade degree is tested.Simulation results show that comparison with current image reconstruction techniques,this algorithm has ideal reconstruction quality,and the better user response of this algorithm is obtained when the image degradation degree is large.