基于空间约束外点处理的图像非盲去卷积

Non-blind image deconvolution base on handling outliers spatially constrained

提出一种鲁棒的基于空间约束外点处理的图像非盲去卷积方法.借助马尔科夫随机场,构造图像邻域像素点的空间约束信息,对模糊图像的内点和外点分别建模,结合空间约束信息,提出一种新的估计潜在清晰图像的目标函数,并通过最大后验概率估计来求解潜在的清晰图像.该方法可以直接有效地利用EM(expectation-maximization)迭代算法来优化参数.实验结果表明,提出的方法有效、鲁棒,在抵抗噪声、亮度变化等方面表现出良好的性能.

A novel robust non-blind image deconvolution method based on handling outliers spatially constrained was proposed. Based on Markov random field, incorporating spatial constrained information between neighboring pixels for deconvolution. A novel objective function of underlying image estimating upon spatially constrained information was proposed, which explicitly took inlicrs and outliers into account. The most probable underlying image can be formulated as a maximum a posteriori estimation problem of the objective function. The proposed method could directly apply the EM （expectation-maximization） algorithm to optimize the parameters. Experimental results demonstrate that the proposed method is robust and effective, and shows good performance in noise resistance and brightness change.