Image denoising and enhancement is one of the most important tasks in image processing. Self-snake model is a most basic and important image denoising method based on PDE (Partial Differential Equation), in which the stop function is very important and is usually related to the gradient of the image. But it?s difficult to choose the smoothness scale as we usually do in acquiring the gradients of the approximating image by smoothening it with a single scale. So we proposed a multiscale gradient vector which integrate the different scale gradient vector from coarse image to fine according to our some defined rulers instead of single-scaled gradient vector of images to improve the performance of self-snake model. The efficiency of our proposal was proved by experimental results.