For the noise amplification problem during gradient transform based image enhancement, an improved algorithm combined with Gaussian curvature filter was proposed. First, the nonlinear gradient transformation is utilized for image gradient field to enhance the contrast. Then the enhanced image is reconstructed by minimizing an energy functional through gradient descent method. During iteration of gradient descent, the reconstructed image and its first and second derivatives are smoothed by Gaussian curvature filter, which can solve the promblems of nonlinear amplification and diffusion of noise in image reconstruction and preserve details. The experimental results with various images show the proposed algorithm can effectively suppress noises and enhance the image edges contrast at the same time.