High-quality multispectral imaging based on edge priors

In order to more effectively preserve the high-frequency information in multispectral demosaicing image processing, this paper proposes a novel high-quality multispectral demosaicing method based on edge priors. This multispectral demosaicing method aims to restore the image by reconstructing the values of all unsampled bands at each pixel position for the raw images generated by such a multispectral filter array (MSFA). First, the system arranges nine bands in a 4x4 repeating pattern, with the dense band occupying half the space and the other bands each occupying 1/16 of the space. Then, it utilizes differences in directional gradients between neighbor bands to guide the demosaicking process of the dense band. Subsequently, the reconstructed dense band is used as a guide image, and other bands are reconstructed using guided filtering and residual interpolation techniques to achieve more accurate reconstruction results. Experimental results show that the proposed method has significantly improved over the current popular demosaicing technology for nine-band multispectral imaging in terms of Peak Signal-to-Noise Ratio (PSNR), Structural Similarity Index (SSIM), and Spectral Angle Mapper (SAM). These results demonstrate that the proposed method can better preserve the local structure and edge information of the original image, effectively reduce the generation of edge artifacts, and thus significantly enhance the reconstruction quality and accuracy of multispectral images.