Image Denoising: A Multi-Scale Framework Using Hybrid Graph Laplacian Regularization

Abstract

In this paper main aim is to focus on to remove impulse noise from corrupted image. Here present a method for removing noise from digital images corrupted with additive, multiplicative, and mixed noise. Here used hybrid graph Laplacian regularized regression to perform progressive image recovery using unified framework. by using laplacian pyramid here build multi-scale representation of input image and recover noisy image from corser scale to finer scale. Hence smoothness of image can be recovered. Using implicit kernel a graph Laplacian regularization model represented which minimizes the least square error on the measured. A multi-scale Laplacian pyramid which is framework here proposed where the intra scale relationship can be modelled with the implicit kernel graph Laplacian regularization model in input space inter scale relationship model with the explicit kernel in feature space. Hence image recovery algorithm recovers the More image details and edges