基于先验信息的压缩感知图像重建方法分析-analysis of compressed sensing image reconstruction method based on prior information.docxVIP

摘 要 II Abstract III 分析图像块稀疏编码系数的先验，用 Laplacian 分布对其进行建模，同时考虑图像 编码中的噪声，定义稀疏编码系数的保真项；通过迭代方法估计 Laplacian 分布的 中心与稀疏编码系数。在第四章方法的基础上提出了一种基于图像块结构与稀疏 系数双重先验信息的压缩感知重建方法，从系数自身分布及抑制重建图像与原始 图像系数之间误差出发，为压缩感知重建模型增加新的先验正则项。对常见的自 然图像进行实验，结果表明，此方法比传统的压缩感知重建模型在评价指标方面 有较大的提高，在视觉效果方面更加清晰。 本文的工作得到了国家重点基础研究发展计(2013CB329402,NCET-10-0668), 国家自然科学基金60971112和高等学校学科创新引智计划 (111 计划)：No. B0704)的资助。 关键词：压缩感知 先验信息 结构字典 统计先验 双重先验  Abstract Compressed sensing (CS) is a new sampling framework for acquisition and processing of signals. The main idea of CS is that signals can be accurately or approximately reconstructed through non-related measurement of sampling data in low dimension based on the sparsity or compressibility of signals. Therefore, compressed sensing can sample and compress signals at the same time, and make the sampling and processing with low cost possible. Compressed sensing opens the door for the traditional signal processing theory built on Nyquist sampling theorem from a new viewpoint, and gains wild attention of Math, Signal Processing, Image Processing and Machine Learning and so on. At present, the researches on compressed sensing are developed from early concept understanding, numerical simulation, theory validation and system designing to deepen the theory, as well as developing and applying of actual system. The reconstruction of signals at low sampling rate is at the cost of solving a nonlinear optimization problem under the theory of compressed sensing, and the quality of the solution influences the quality of reconstructed signals, and then determines the availability and utility of compressed sensing technology. Most of the existing models of signal reconstruction only utilize the sparsity or compressibility of signals. However, the actual signals themselves, like images, have lots of valuable prior information, such as local structural information, self-similarity, except for sparsity. Exploiting the prior information and taking advantage