散焦含噪图像的点扩散函数估计与亚像素边缘检测

针对机器视觉中图像模糊的原因，采用广义高斯函数描述点扩散函数(PSF)，建立边缘过渡区模型，从一幅图像所含多条边缘中提取样本，依据提出的整合准则从全部样本的估计结果中计算整幅图像的PSF，利用最优求解的概念，可实现最小二乘意义下PSF及亚像素级边缘位置的最佳估计。实验结果表明, 由于算法获得图像所对应PSF的最佳估计，边缘检测具有比较高的抗噪性能，在边缘亚像素检测基础上完成的尺寸测量误差小于0.5%。

PSF Estimation & Sub-pixel Edge Detection for Defocused Noisy Image

Considering the blur of images in machine vision, the generalized Gaussian function was selected to describe the point spread function (PSF). Firstly a edge spread model was built. Then some samples belonging to several edges in an image were obtained. The two parameters of PSF corresponding to every sample were estimated. In order to integrate all the results of every samples, a integration rule was proposed. Utilizing the idea of optimal solution, the optimal PSF estimation and sub-pixel level edge detection were achieved. The experiment results showed that the better immunity from noise of the proposed algorithm benefited from the accurate estimation of the PSF. The error of dimension measuring based on the sub-pixel level edge detection was less than 0.5%.