改进的奇异值分解法估计图像点扩散函数

Estimation of PSF of image system using modified SVD method

为了提高图像复原算法的性能,提出了一种改进的奇异值分解法估计图像的点扩散函数。从图像的退化离散模型出发,对图像进行逐层分块奇异值分解,并自动选取奇异值重组阶数以减少噪声对估计的影响。利用理想图像奇异值向量平均能谱指数模型,估计点扩散函数奇异值向量的频谱,再反傅里叶变换得到其时域结果。实验结果表明,该方法能在不同信噪比情况下估计成像系统的点扩散函数,估计结果比原有估计方法有所提高,有望为图像复原算法的预处理提供一种有效的手段。

To improve the performance of image restoration algorithms, a modified Singular Value Decomposition（SVD） method was proposed to estimate the Point Spread Function （PSF） of an imaging system. Using the discrete image degradation model, a block-based SVD filter scheme was applied for the image denoising with an automatically determined singular value rank. After the spectra of PSF singular vectors were estimated under an exponential model for the averaged spectra of un-degraded image singular vectors, the IFFT was used to get the time-domain estimation of the PSF. The experimental results show that this proposed method can be applied to estimate the PSF of the imaging system under a wide SNR range and its performance is better than the original method. It may be used as an effective method for the image preprocessing in image restoration problems.