An Optimal Weight Model for Single Image Super-Resolution
Dinh Hoan Trinh, Marie Luong, Jean-Marie Rocchisani, Canh Duong Pham, Françoise Dibos, Huy Dien Pham · 2012
In this paper, a novel example-based super- resolution method is introduced. The objective is to estimate a high-resolution image from a single low- resolution image. By considering an image as a set of small image patches, our method is performed on each patch with the help of a given database of high and low-resolution image patch pairs. For each given low-resolution patch, its high-resolution version is considered as a sparse positive linear combination of the high-resolution patches from the database. The coefficients of this combination are referred to as the weights, and an optimal weight model is proposed to find this combination such that the high-resolution patch is consistent with the low- resolution patch while being similar to the best candidate high-resolution patches from the database. Experimental results show the good performance of our method over some state-of-the-art methods and confirm the efficiency of the proposed method.