Abstract— in this paper main aim is to focus on to remove impulse noise from corruptedimage. Here present a method forremoving noise from digital images corrupted with additive, multiplicative, andmixed noise. Here used hybrid graph Laplacian regularized regression toperform progressive image recovery using unified framework. by using laplacianpyramid here build multi-scale representation of input image and recover noisyimage from corser scale to finer scale. Hence smoothness of image can be recovered. Using implicit kernel a graph Laplacianregularization model represented which minimizes the least square error on themeasured. A multi-scale Laplacian pyramid which is framework hereproposed where the intra-scale relationship can be modelled with the implicitkernel graph Laplacian regularization model in input space inter-scalerelationship model with the explicit kernel in feature space. Hence imagerecovery algorithm recovers the Moreimage details and edges
Keywords:- impulse noise, graph laplacian regularized regression, multi-scaleframework.