Analysis on Weighted Average Between Features in Dictionary Learning and Sparse Representation Algorithms for Low-Resolution Images

Abstract

Recently, the problem of Low-Resolution (LR) is happened to be the key challenge in the field of image processing. To tackle this problem, Super-Resolution (SR) techniques have been developed. Generally, SR is used to acquire more information about an image by recovering an HR image from the LR image without losing high frequency details [1]. This paper is focused on analysing the effectiveness of using combined features in dictionary learning and sparse representation algorithms for producing images with better resolution. The method used to combine the properties of features: energy (F1) and entropy (F2) extracted in this paper is known as the weighted average techniques. In this case, different combination of weightage which was written as [W1, W2] will be assigned to F1 and F2 respectively. As a result, the weightage combination of [2, 8] achieved the highest improvement in PSNR values of 6.0863 dB and the second highest improvement in SSIM values of 0.2559. In conclusion, the SR system constructed based on the dictionary learning and sparse representation algorithms with the use of weighted average between features is able to solve the image LR problems. This work can be improved by testing on more input images obtained from databases.KeywordsImage processingDictionary learning and sparse representation