A segmentation method based on texture gradient is proposed for the high resolution remote sensing images which have rich texture information.First, Nonsubsampled Contourlet Transform is used to extract texture features.Second, texture gradient is calculated, and then it is combined with the intensity gradient to get the final gradient.At last, the labeling watershed algorithm is used in segmentation with the fmal gradient.This method not only avoids the over segmentation problem, but also preserves the edge information.The experimental results over aerial image and Quickbird satellite image show that the proposed method is well-behaved and efficient on high resolution remote sensing images.