Iteratively merging information from a pair of flash/no-flash images using nonlinear diffusion

In this paper, we present a mechanism based on the guided filter for merging information from a pair of imperfectly exposed images, to produce a better one. The guided filter [7] utilizes information from a guide image to enhance a corrupted input image as similarly done in the joint bilateral filter [3, 12]. Here we analyze the spectral behavior of the guided filter kernel in a matrix formulation and introduce a novel iterative technique which significantly improves it. Iterations of the proposed method consist of a combination of nonlinear anisotropic diffusion and reaction-diffusion. We demonstrate that the proposed approach outperforms state of the art methods in both flash/no-flash restoration where one of the images is significantly noisy or blurry.