Adaptive iterative spectral correlation-based multispectral demosaicking using pseudo-panchromatic image

Abstract Multispectral image demosaicking is a key technique for recovering full-resolution multispectral images from sparsely sampled data acquired by a multispectral filter array. To address the fixed guidance image of existing methods ignoring the spectral correlation (SD) between different bands, this paper proposes an adaptive iterative SD-based multispectral image demosaicking method using a pseudo-panchromatic image (PPI). First, the method generates a PPI from the mosaic image via weighted smoothing filtering, and recovers the initial multispectral image by applying the guided filtering method. Second, it quantifies the SD between different bands using the Pearson correlation coefficient and constructs the band-selected PPI to estimate the multispectral image. Finally, it performs an adaptive iterative strategy with dynamically updated band-selected PPI to adaptively select the optimal parameters, thereby achieving the joint optimization of spatial details and spectral fidelity. Experimental results show that the proposed method outperforms existing methods in both subjective and objective evaluation metrics, providing an effective solution for multispectral image demosaicking.