Abstract: Distance regularized level set evolution model for image segmentation has some shortcomings. Initial curve is not flexible and the direction of curve evolution is not self-adaptive. A new active contour model was proposed. A signed pressure force function (SPF) was constructed, integrating local information and global information, and shown to modulate the signs of pressure forces inside and outside the region of object; Then SPF was embed to DRLSE model to solve the problem of requiring the initial contour to surround (let in or keep out) the objects. Experimental results show that the new model's initial contour can be placed anywhere in the image, the new model can effectively segment images with intensity inhomogeneity and weak boundaries and it can effectively segment images containing multi-objects or deep recessed area.

Key words: active contour, DRLSE model, SPF function, local information, global information