In H.264 standards, video coding is carried out in two steps. First, pixels are predicted by copying previously reconstructed known neighbour pixels along an angular direction inside the block. Then, the difference between original and predicted is transform-coded with 2-D Discrete Cosine Transform (DCT). Each block pixel is predicted from only few directionally aligned neighbour pixels of the block. Though it is a useful approach, it eliminates some useful neighbour pixels of the block. To use this, a linear prediction method is proposed, where each block pixel is predicted using a weighted sum of top row and first column neighbour pixels of the block. The drawback of this method is the increased complexity because of the large number of pixel usage. In this paper, we propose a another method to intra prediction, where we predict image pixels using a 2D Markov process. We show that combining the intra prediction approach and the LWT transform improves overall compression performance. In this paper, both the intra prediction and the transform steps are obtained based on 2D Markov processes.