Multi-frame demosaicing for the Quad Bayer CFA in the color difference domain

Area of Science:

• Image processing
• Computational imaging
• Computer vision

Background:

• Quad Bayer color filter arrays (CFAs) offer high sensitivity in low light via pixel binning.
• Full-resolution imaging with Quad Bayer CFAs requires direct demosaicing under sufficient light.
• Quad Bayer CFAs exhibit increased aliasing artifacts compared to Bayer CFAs due to structural limitations.

Purpose of the Study:

• To address aliasing artifacts in Quad Bayer color filter array (CFA) imaging.
• To propose a novel multi-frame demosaicing method for enhanced spatial and color resolution.
• To improve the performance of demosaicing algorithms for submicron image sensors.

Main Methods:

• A multi-frame demosaicing approach is proposed.
• Local spatial contributions within windows are assessed to reconstruct resolution lost to aliasing.
• The color difference domain is utilized to exploit inter-channel correlations, reducing color artifacts and enhancing resolution.

Main Results:

• The proposed multi-frame demosaicing method effectively reconstructs spatial resolution.
• Color artifacts are alleviated, and spatial resolution is enhanced by exploiting inter-channel correlations.
• Experimental results show superior quantitative and qualitative performance compared to single-frame and pixel-binning multi-frame methods.

Conclusions:

• The proposed multi-frame demosaicing method effectively mitigates aliasing artifacts in Quad Bayer CFAs.
• The technique significantly enhances both spatial and color resolution in full-resolution imaging scenarios.
• This method offers a substantial improvement over existing demosaicing strategies for advanced imaging systems.