Aiming at these problems that image segmentation model oriented on level set model can not accurately segment object and is sensitive to noise and initial contour, this paper proposed an active contour model based on joint constraint with superpixel/pixel and sparse decomposition. First, it introduced the superpixel to extract image block information and constructed a symbol pressure function to prevent contour evolution from falling into local optimization. Then it constructed the energy function relied on superpixels and pixels to remedy the defect that the superpixels couldn't retain local detials. In order to solve the problem of slow evolution speed for active contour model based on non-global information, it utilized superpixels to accelerate the contour evolution. Finally, this paper introduced the constraint conditions based on sparse decomposition to weaken the influence of local noise. Compared with various level set methods, the experimental results verify the effectiveness of the proposed method. The proposed method is more robust to noise and initial contour location, the average Jaccard similarity coefficient is increased by 34% compared with original selective binary and Gaussian filtering regularized level set.

[2021-03-05] Printed Article /doc/2019.12.0697/v2