To solve the problem that active contour model based on image edge information is sensitive to initial contour, a new adaptive level set evolution model combining local entropy gradient is proposed. The local entropy gradient and regularized image gradient are used to construct the area term weight coefficients to replace the original constant coefficient.  The novel weight coefficients can change the symbol adaptively with the image information so that the evolution function adaptively moves up and down according to the image information, allowing constant initialization. The proposed model solves the initial contour sensitivity problem and allows constant initialization of the level set. Experiments verify that the model has good segmentation effect on gray uneven image, noise image, partial multi-object image and subjective edge image. In addition, the proposed adaptive weight coefficients can be used in other segmentation models which involve  area energy terms.