论文部分内容阅读
为了对图像复原算法频谱恢复特性进行分析和评价,提出了一种基于高斯函数假设的分析新方法。该方法假设光学传递函数H和退化图像频谱函数G为高斯函数,采用方差以及提出的方差比作为频谱宽度指标,对图像复原算法的频谱恢复特性进行定量分析和评价。分析中对H和G曲线设定两组方差,分无噪声和有噪声两种情况,计算出约束最小平方滤波法(CLS)和最大似然法(PML)等图像复原算法复原的图像频谱曲线及其方差和方差比,采用计算结果对复原算法进行定量的分析和评价,获得良好的效果。分析指出,最大似然法的频谱外推能力和噪声抑制特性均明显好于约束最小平方滤波法。对两种算法的分析评价实验表明,高斯函数假设分析方法是一种简便有效的图像频谱恢复特性分析方法。
In order to analyze and evaluate the spectral recovery characteristics of image restoration algorithm, a new analysis method based on Gaussian function hypothesis is proposed. The method assumes that the optical transfer function H and the degraded image spectral function G are Gaussian functions, and the variance and the proposed variance ratio are used as the indicators of spectral width to quantitatively analyze and evaluate the spectral recovery characteristics of the image restoration algorithm. In the analysis, two groups of variance of H and G curves were set, and no noise and no noise were used to calculate the image spectral curve reconstructed by image restoration algorithms such as constrained least squares (CLS) and maximum likelihood (PML) And its variance and variance ratio, using the calculation results of the recovery algorithm for quantitative analysis and evaluation, and achieved good results. Analysis shows that the maximum likelihood method of spectral extrapolation ability and noise suppression characteristics are significantly better than the constrained least squares filter. The analysis and evaluation experiments of the two algorithms show that the Gaussian function hypothesis analysis method is a simple and effective method of spectrum spectral characteristics analysis.