Image Super Resolution and Denoising Using Block Diagonal Representation

Abstract

Single image super resolution is the procedure of achieving a high resolution image from a low resolution image without any loss in the textural and structural information. First we apply the rational fractal interpolation model in the texture and non-texture region. Next, we develop an algorithm based on the proposed model. The scaling factor in the texture region is to be calculated. Even though our algorithm achieves a competitive performance with finer details, enhancing the sparsity will give a better result. Multispectral image denoising is attained using block diagonal representation. We generalize patch level representation via block diagonal matrix, investigate the proper choice of patch level transform and modifies through Single Value Decomposition(SVD). We demonstrate that a very competitive performance can be obtained through the combination of a proper global patch basis and local PCA.