Adaptive window-based multispectral image demosaicking method using pseudo-panchromatic image

Multispectral image demosaicking is a challenging problem for single-sensor multispectral imaging systems with the multispectral filter array. In this paper, we propose an adaptive window-based multispectral image demosaicking method using a pseudo-panchromatic image (PPI). This method aims to use the spectral correlation between channels to recover the missing pixels in the mosaic image. First, a set of weighted average filters based on adaptive sparse windows is used to generate a PPI from the mosaic image. Then, the guided filtering window is adaptively adjusted according to the differences in the number of sampled pixels in each channel of the mosaic image, with the PPI as a guide to generate the initial multispectral image. Finally, the initial multispectral image is used to generate a more correlative local spectral average image, which guides the generation of high-quality multispectral images. Experimental results show that the proposed multispectral image demosaicking method outperforms existing methods in both subjective and objective evaluations.