The image denoising methods based on Partial Differential Equations (PDEs) were explored. In order to alleviate the staircase effect in second-order PDE (P-M model) and improve the ability of edge and texture preserving of fourth-order PDE (Y-K model), the Gradient Vector Convolution (GVC) field was introduce into the fourth-order PDE, and a four-order anisotropism diffusion model was established. Firstly, the parts of diffusion in the direction of gradient was subtracted. Then, the GVC field was introduced to replace the calculation of second derivative. Because of the robustness of GVC and its outstanding ability of detecting edge, an effective anisotropic diffusion model was obtained. Experimental results indicate that the GVC based fourth-order model can protect the details over the original model like edge and texture features better and can improve the Peak Signal to Noise Ratio(PSNR).The PSNRs in experiments have been improved more than 1 dB as compared with that original Y-K model.