基于模糊结构图的模糊核估计

发布时间：2018-07-18 14:05

【摘要】：图像结构边缘对模糊核估计有重要意义.近年来许多成功的算法都致力从潜在清晰图像中分离出结构边缘形成中间图像,然后用其与模糊图像一起估计模糊核.但是这些算法忽视了从模糊图像中分离出结构边缘对应的部分,导致核估计过程中目标函数的数据项不平衡.针对这一问题,本文利用中间图像和潜在模糊核产生二值模板对模糊图像进行处理,分离出结构边缘对应的部分,并用其修正目标函数.此外本文提出采用L0范数同时约束幅值域和梯度域的正则项,从而缩小核估计的解空间.多个标准测试数据库上实验结果表面,本文算法无论在鲁棒性还是准确性方面均具有更好的效果.
[Abstract]:The edge of image structure plays an important role in fuzzy kernel estimation. In recent years many successful algorithms have been devoted to separating structural edges from potential clear images to form intermediate images and then using them to estimate blur cores together with blurred images. However, these algorithms neglect to separate the corresponding parts of the structure edge from the blurred image, which leads to the imbalance of the data items of the objective function in the kernel estimation process. To solve this problem, the binary template generated by the intermediate image and the potential fuzzy kernel is used to process the blurred image, and the corresponding part of the structure edge is separated, and the objective function is modified by it. In addition, it is proposed that L _ 0 norm is used to simultaneously constrain the regular terms in both amplitude and gradient domains, so that the solution space of kernel estimation is reduced. On the surface of experimental results on several standard test databases, the proposed algorithm is more effective in both robustness and accuracy.
【作者单位】：合肥工业大学计算机与信息学院;光电控制技术重点实验室;中国科学技术大学自动化系;
【基金】：国家自然科学基金(No.61175051,No.61175033) 安徽省自然科学基金(No.1508085SMF222) 光电控制技术重点实验室和航空科学基金联合资助项目(No.201451P4007)
【分类号】：TP391.41

【相似文献】