MAP based resolution enhancement of video sequences using a Huber-Markov random field image prior model

In this paper, we propose two approaches for video sequence resolution enhancement using maximum a posteriori (MAP) estimation. Huber-Markov random fields (HMRF) are used as prior models. These models can better preserve image discontinuities (edges) when compared with Gaussian prior models. The two proposed approaches differ in the selection of image smoothness measure. The first approach employs a measure that is based on a discrete Laplacian kernel, while the second approach uses a finite difference approximation of second order derivatives at each pixel of the high-resolution image estimate. Experimental results are presented and conclusions are drawn.