In order to make the fused remote sensing image have excellent contrast while highlighting the texture features, a double restriction model to fuse the remote sensing image in the non subsampling shearlet transform domain is designed. Firstly, the multispectral image is analyzed by IHS model, and divided into intensity, hue and saturation components. Then, with the help of NSST, the low and high frequency coefficients of I component and panchromatic image are analyzed. Finally, through the information entropy and mean value model, the image rich information and brightness richness are calculated, and the low-frequency fusion coefficient is calculated. Using the convolution operation of image and direction matrix, the texture features of image are calculated. By calculating the standard deviation of image, the contrast information of image is obtained. Using the texture and contrast information of image, the double restriction model is constructed, and the high-frequency fusion coefficient is calculated with it, and then the fusion image is obtained. The experimental results show that the fusion performance of this algorithm is better than the existing algorithm, and the fusion image texture is more prominent, the contrast is better, and the information richness is higher.