关于柯西噪声去除模型的自适应参数的研究.pdf

摘 要 在柯西噪声的变分模型中平衡参数 控制着正则项与保真项之间的权衡, 起 着⾮常重要的作⽤，⼀个全局的约束并不能产⽣理想的效果。如何根据噪声和图 像的特性, 提出局部约束下柯西噪声的变分模型，实现平衡参数的⾃动迭代是 本⽂研究的问题。本⽂⾸先介绍柯西噪声的变分模型，然后提出局部约束下柯 西噪声的变分模型, 并证明解的存在性; 最后提出空间⾃适应的全变分柯西噪声 去噪算法 SA-TV-C (Spatially Adaptive Total Variation for Cauchy noise) 和改 进的 SA-TV-C 算法。实验结果表明 SA-TV-C 算法⽐交替⽅向乘⼦法 ADMM (Alternating Direction Method of Multipliers) 恢复的图像质量更好。相⽐ SA- TV-C 算法, 改进后的SA-TV-C 算法恢复的图像质量得到进⼀步提升。 关键词: 变分模型, 局部约束, 局部保真项, 柯西噪声, ⾃适应参数选择 i ABSTRACT In the variational model of Cauchy noise, the balance parameter controls the trade-off between the regular term and the fidelity term, and plays a very important role. A global constraint can not produce a good effect. According to the characteristics of noise and image, how to propose the variational model of Cauchy noise under local constraints and realize the automatic iteration of balance parameters is the problem studied in this paper. In this paper, the variational model of Cauchy noise is introduced firstly, then the variational model of Cauchy noise under local constraints is proposed, and the existence of solutions is proved; finally, the spatial adaptive total variation for Cauchy noise algorithm SA-TV- C (Apatial Adaptive Total Variation for Cauchy noise) and the improved SA- TV-C algorithm are proposed. The experimental results show that SA-TV-C algorithm has better image quality than ADMM (Alternative Direction Method of Multipliers). Compared with SA-TV-C algorithm, the image quality of the improved SA-TV-C algorithm is further improved. Keywords: Variational model, Local constraint, Local fidelity term, Cauchy noise, Adaptive parameter selection iii 目 录 摘 要