基于奇异值分解的性质,从离散退化模型出发,采用理想图像奇异值向量的平均能谱理论,得出用奇异值分解来估计点扩散函数的方法,复原过程用逆滤波法来实现.文中采用奇异值累计和函数的二阶导来确定点扩散函数估计过程中的重组阶数R1,三阶导曲线用于去噪的重组阶数R2的选取.用本文方法进行去噪复原实验,与自动选取法进行比较,去噪复原效果较好.%Based on the discrete image degrading model, we derived the amplitude estimation equations of the first order sigular value vecters of the degraded point spread function (PSF) .The spectra of PSF singular vectors were estimated under an exponential model for the averaged spectra of undegraded image singular vectors. Restoration of the degraded image was achieved by inverse filtering.The second derivative of the cumulative sum function of sigular values was proposed to determine the rank Rl for PSF estimation, while The third one used to choose the rank R2 for de-noising.As experience proves, It is better than the automatic selecting method.
展开▼