Image Super-Resolution: Use of Self-learning and Gabor Prior

Abstract

Recent approaches on single image super-resolution (SR) have attempted to exploit self-similarity to avoid the use of multiple images. In this paper, we propose an SR method based on self-learning and Gabor prior. Given a low resolution (LR) test image I ₀ and its coarser resolution version I _− 1, both captured from the same camera, we first estimate the degradation between LR and HR (I ₁) images by constructing the LR-HR patches from LR test image, I ₀. The HR patches are obtained from I ₀ by searching for similar patches (of I ₀) of the same size in I _− 1. A nearest neighbor search is used to find the best LR match which is then used to obtain the parent HR patch from I ₀. All such LR-HR patches form self-learned dictionaries. The HR patches that do not find LR match in I _− 1 are estimated using self-learned dictionaries constructed from the already found LR-HR patches. A compressive sensing-based method is used to obtain the missing HR patches. The estimated LR-HR pairs are used to obtain the LR image formation model by computing the degradation for each pair. A new prior, called Gabor Prior, based on the outputs of a Gabor filter bank is proposed that restricts the solution space by imposing the condition of preserving the SR features at different frequencies. The experimental results show the effectiveness of the proposed approach.