基于模糊图像边界梯度变化的模糊核估计研究

Research on estimation related blur kernel using the gradients of boundaries in blurred images

通过分析模糊图像边界的梯度变化,提取了导致图像模糊的运动特性信息,由此建立了与运动特征信息相关的模糊核初始化模型,在相同的恢复质量下,提高了去模糊算法效率。实验表明,该算法提取出了高效的初始化模糊核,降低了对后期递归优化算法的依赖,去模糊耗时短,效果明显,尤其适用于平坦背景、边界区域变化明显的模糊图像去模糊。

This paper analyzes the gradients of boundaries in blurred images, and extract the motionfeatures which lead to blur. Then the blur kernel related to motion information is modeled and initialized.The efficiency is improved in case of the same recovery effect. The experimental results show that theproposed initialized blur kernel weaken the dependence of recursion optimization in the process ofdeblurring. The cost time reduced and the effects are good especially for images with flat background andsignificant changes near boundaries.