A Fast Image Denoising Method Based on CP Tensor Analysis

The data represented by tensor can maintain its original form, which is conductive to preserving high dimensional structure and adjacent relation information of data. Tensor factorization plays an instrumental role in signal processing, especially image processing. However, the existing tensor-based fitting algorithms require a lot of time cost and have low fitting accuracy, which is more obvious when the image data is large. In this paper, an accelerated fitting algorithm based on the Canonical Polyadic (CP) tensor analysis is designed to process images. Compared with the existing tensor-based fitting algorithm, the developed algorithm greatly improves the fitting speed with high fitting accuracy, and can quickly reduce the noise of the image. Simulation results for color image show that our method have greatly improved the efficiency of image denoising.