Remote sensing images often suffer from stripes,which degrades the image quality. To remove stripes,a new destriping method based on spatial adaptive total variation is proposed. Firstly,the stripe model is built and the gain and bias of the model are estimated with high SNR regions. Then,an energy function is designed under the framework of total variation and an adaptive regulation parameter is introduced,which adapts to the spatial varying information. At last,the split Bregman approach is exploited to optimize the minimization problem. Experiment results demonstrate that the proposed method can improve the equivalent number of looks (ENL) from 37.26 to 76.48,and the improvement factor (IF) of radiometric quality is improved to 8.52 dB. The proposed method can effectively remove stripes and preserve detail information.