紧缩感知及其在图像处理中的应用.pdf

压缩感知及其在图像处理中的应用 摘 要 本文系统地介绍了压缩感知这一信号处理技术中新兴领域的有关基 本概念，分析了现有的信号重构方法的优缺点。根据二维图像小波变换 系数层的特点和现实应用中稀疏度未知的情况，提出了基于单层小波变 换的自适应压缩感知新算法，保留图像低频系数，只针对高频系数进行 测量。详细地说，在图像重构时，利用稀疏度自适应匹配追踪算法对高 频系数或逐行或逐列或整体进行恢复，再进行小波逆变换重构图像。仿 真结果表明，与单层小波变换的非自适应压缩感知算法相比，解决了稀 疏度未知情况下的图像恢复问题，且重构图像质量也得到很好的保证。 在相同的采样率下，PSNR 相差不过2dB 。 关键词：压缩感知；稀疏度；小波变换；稀疏度自适应匹配追踪 Compressed Sensing and Its Applications in Image Processing Abstract In this thesis we systematically introduce the related basic concepts about compressed sensing, a newly-developing field in signal processing techniques and analyze the advantages and disadvantages of the current algorithms for the signal reconstruction. According to the properties of wavelet transform sub-bands and the unknown sparsity in many practical applications, an improved adaptive compressed sensing algorithm based on the single layer wavelet transform was proposed, which only measured the high-pass wavelet coefficients of the image but preserving the low-pass wavelet coefficients. For the reconstruction, by using sparsity self-adaptive matching pursuit (SAMP) algorithm, high-pass wavelet coefficients could be recovered by row, column or whole of the measurements. Then the image could be reconstructed by the inverse wavelet transform. Compressed with the original compressed sensing algorithm based on the single layer wavelet transform, simulation results demonstrated that the proposed algorithm solve the problem of image reconstruction in the case of unknown sparsity, and the quality of the recovered image is guaranteed. For the same sampling rate, the PSNR difference of the proposed algorithm and the original algorithm is not greater than 2dB. KEY WORDS ：comp