Abstract: Aiming at the fact that the DRLSE level set model is inadequately sensitive to noise and dependent on the initial contour and slow evolution we used wavelet transform and wavelet threshold denoising methods. A new edge stop function and adaptive weight coefficient based on image information are defined by constructing the edge characterization matrix which is not sensitive to noise. An improved DRLSE level set image segmentation model is thus obtained. The finite difference method is employed to solve the model, and Jaccard similarity is used as the quantitative analysis method of evaluation model. The numerical results show that the improved model and algorithm are effective for image segmentation, overcoming the limitation of DRLSE level set model and dividing the noisy image and defining the initial contour position, which improve the computational efficiency and image segmentation precision of the DRLSE level set model.

Key words: image segmentation, DRLSE level set, edge stop function, adaptive