Image Block Compression Based on Dual-Learning Dictionaries

Abstract

To improve the nonlinear reconstruction performance of sparse representation in image compression and reconstruction process, a new image compression algorithm based on dual synthesis and analysis dictionaries is proposed. Firstly, the dual synthesis and analysis dictionaries learning algorithm is analyzed, including penalty function analysis and the algorithm of dictionary learning with gradient decent. Then, the compression model based on synthesis and analysis dictionaries is introduced and the properties of the model is also analyzed. Experimental results demonstrate that the proposed method outperforms many other algorithms in image compression.