A dictionary-learning algorithm based on method of optimal directions and approximate K-SVD

Abstract

Dictionary design as a fundamental ingredient in the Sparse-Land model is an important problem and has attracted increased attention. A proper dictionary can increase sparsity levels for sparse presentation and lead to better performance in the various applications. There are many effective algorithms proposed for the task of dictionary learning, such as the Method of Optimal Directions (MOD) algorithm and the K-Singular Value Decomposition (K-SVD) algorithm. The K-SVD algorithm presents performance better than the MOD algorithm, but the higher computational complexity of updating the atoms restricts its application. In this paper, we propose a new method that combines MOD with Approximate K-SVD (AK-SVD) for dictionary learning and show that it converges faster than K-SVD while obtaining better results. Experimental simulations demonstrate the efficiency of our proposed algorithm and its promising performance on the recovery of a known dictionary and dictionary learning for natural image patches. We also implement our proposed algorithm in the image denoising and evaluate its image denoising performance.