Linear mixture model-based image fusion

Traditional remote sensing image fusion is based on spectral space. When adding texture information of a high-resolution image to a low-resolution image, the original spectral information of low-resolution image would be distorted more or less. In this paper, a schema to set up a linear relationship between a low-resolution image and a high-resolution image in a different feature space was explored. And then, a look up table based on that relationship was applied to unmix mixture pixels in the low-resolution image respectively. An example was illustrated of unmixing and fusion of the multispectral reflective bands and panchromatic band of a Landsat ETM+ image in Shunyi district, Beijing and obtained a satisfactory fusion result.