两次引导滤波的显微视觉散焦图像快速盲复原

发布时间：2019-08-13 19:52

【摘要】：针对显微图像盲复原算法存在的计算量大、振铃效应以及噪声敏感的问题,提出贝叶斯框架下两次引导滤波的快速盲复原算法。利用显微图像成像原理中基于深度信息估计点扩展函数的概率模型,构建了贝叶斯框架下盲复原的最小优化问题;通过分析最大后验概率的最小优化问题求解过程,推出了实施引导滤波器可快速求解优化问题的结论;为有效去除振铃和噪声,设计了两次引导滤波的求解方案,其将第一次引导滤波求解的结果作为优化问题的二次输入。实验结果表明,复原结果的像素误差率约为0.04,较常用盲复原算法的复原准确度提高了约20%,运行时间也大幅缩短,该方法能有效应用于显微视觉下微装配散焦图像盲复原的工程实践中。
[Abstract]:In order to solve the problems of large amount of computation, ringing effect and noise sensitivity in the blind restoration algorithm for microscopic images, a fast blind restoration algorithm based on two guided filtering is proposed. Based on the probability model of estimation point spread function based on depth information in microscopic image imaging principle, the minimum optimization problem of blind restoration in Bayesian framework is constructed, and the conclusion that the implementation of guided filter can solve the optimization problem quickly is derived by analyzing the solving process of the minimum optimization problem with maximum posterior probability. In order to effectively remove ringing and noise, a scheme of twice guided filtering is designed, which takes the result of the first guided filtering as the secondary input of the optimization problem. The experimental results show that the pixel error rate of the restoration result is about 0.04, which is about 20% higher than that of the common blind restoration algorithm, and the running time is also greatly shortened. This method can be effectively applied to the engineering practice of blind restoration of micro-assembly defocus image under microscopic vision.
【作者单位】：西南财经大学经济信息工程学院;四川师范大学工学院;宁夏大学信息工程学院;
【基金】：国家自然科学基金重大项目(91218301);国家自然科学基金青年基金(61502396) 宁夏自然科学基金(NZ15054) 西南财经大学中央高校基本科研业务费专项资金(JBK150503);西南财经大学中央高校基本科研业务费青年教师成长项目(JBK170136) 互联网金融创新及监管四川省协同创新中心资助项目
【分类号】：TP391.41

【相似文献】