Design Image Filter Based On Non-local Prior With Applications

Currently,digital image plays an important role in many fields,such as aerospace,mobile payment,transportation,urban security and industrial detection,etc.However,due to the limitations of imaging equipment,environment and transmission condition,digital images are often corrupted by noise,which makes it hard to be applied to the subsequent tasks directly.Image filtering is an important technique to improve image quality,while non-local means filter(NLM)has become a popular algorithm because of its strong detail-preserving property.And more and more traditional local filters have been successfully extended to nonlocal ones,which show better performance compared to the local one.Guided(joint)filters(GIF)has remained an active research topic in designing image filer.The key idea of GIF is to transfer the structural information from a guidance image to an input one.Indeed,by relaxing the dependence on the guidance,static/dynamic(SD)filter avoid texture-copying artifacts effectively.Because of its excellent filtering performance,the static guided filter has attracted more and more researchers' attention.However,due to the locality of its regularizer,SD is prone to another deficiency,i.e.,edge blurring.To solve this problem,in this paper,we extend the static dynamic filter to a nonlocal version(NSD).Firstly,a nonlocal regularizer is established in a subspace transformed by partial least squares,which can better respect the unequal roles of the images.Then,an acceleration approach is designed to reduce the computational complexity induced by the non-local extension,which makes NSD achieve a comparable running time in practice.Thorough experimental results demonstrate that the proposed filter not only can avoid texture copy effectively but also can preserve edges powerfully.In addition,in order to further study the application of non-local prior,we propose a plug-and-play restoration method with an adaptive NLM,and prove its convergence theoretically.