A fast edge detection model combining mixed L1 and L2 fidelity terms

Edge detection plays an immensely important role in image processing. In this paper, we propose a new model with the combination of the L-1 and L-2 fidelity terms on the basis of the well-known Mumford-Shah (MS) model. To solve this minimum model, we design an efficient algorithm based on a fixed-point iterative method and the Split-Bregman (SB) method. Experimental results show that the proposed model and algorithm can get better detected edges and have more advantages in efficiency and accuracy for different pure noisy images, even for mixed noisy images.