Image denoising method based on robust tensor low-rank representation

The invention discloses an image denoising method based on robust tensor low-rank representation, which comprises the following steps: firstly, establishing an objective function of an image noise model, then optimizing the objective function, and solving an equivalent problem after optimization by using a multiplier alternating direction method (ADMM); utilizing a corresponding enhanced Lagrange function and fixing other variables, and respectively and alternately updating epsilon and J in the Lagrange function to solve the optimization problem; when the updating difference value of all variables is smaller than a preset threshold value, outputting a rank representation coefficient tensor. According to the invention, an original noise tensor is decomposed into three parts, namely a recovery data tensor of a low-rank structure, a sparse noise tensor obeying Laplacian distribution and a Gaussian noise tensor obeying Gaussian distribution, and Laplacian noise and Gaussian noise are considered at the same time,