Parallel optimization of K-SVD algorithm for image denoising based on Spark

With the development of society, the image size and resolution are gradually increasing. At the same time, the fast image processing is required. The K-SVD algorithm is an iterative method that alternates between sparse coding and updating the overcomplete dictionary. It is a very simple and highly effective method for the signal of sparse representation. However, the K-SVD algorithm is a computationally intensive algorithm. The traditional computing model will take a lot of time to complete the calculation. As the next generation computing model, Spark not only has a strong computing power, but also has a memory-based fast processing capacity. In order to process image denoising efficiently, this paper proposes a implementation of distributed parallel optimization of K-SVD algorithm (K-SVD-P) on Spark. The results showed that K-SVD-P not only has a good speed-up ratio, but also retains the image texture and other details.