To better maintain coherence of structure and consistency of neighborh ood,this paper proposes an image inpainting algorithm using HSI gradient statistics feature.In the algorithm,th e image firstly is mapped into the HSI color space,then the gradient feature of the image is extracted more fully by use of the HSI gradient.And the statistics such as the mean and variance of the HSI gradient are introduced as parameters to red efine structure similarity between patches,thus the global energy equations of the weighted data term and smooth te rm energy are constructed with it. Meanwhile,the sizes of patches used to compute global energy and the multi-res olution levels are determined adaptively,and the optimization of the global energy is solved by the graph cut s algorithm to obtain the inpainted images.Since the gradient features of the image are extracted more fully in the HSI color space,the new exemplar gradient statistics feature gets more reasonable matching labels,so that image s tructure continuity is better maintained.The experimental results show that the method exhibits good applications and weakens seam effect in both object removal and small missing region in natural images,thus it can keep the consistency of image structure and satisfy human visual effects.