An Improved Injection Model for Pansharpening Based on Weighted Least Squares

Pansharpening is a fusion technique to enhance the spatial resolution of multispectral images by combining with the panchromatic image. This problem can be formulated as detail extraction and injection model, thus the injection estimation is a key for fusion quality. In the literature, the regression-based models are extensively studied, where the solution is usually solved by the ordinary least squares method. To improve the accuracy and robustness of estimation, a new injection model based on weighted least squares is proposed in this paper. The weights are dependent on the local statistics of the input-output pairs, which enhance the regular area and suppress effect of outliers. Experimental results show that the proposed method outperforms the other state-of-the-art methods.