Low-light RGBW Imaging demosaicking method based on residual interpolation prior and a dual-branch decoding network

Area of Science:

• Computer Vision
• Image Processing
• Computational Imaging

Background:

• Traditional Bayer arrays have limitations in spectral range and light intake for low-light imaging.
• RGBW arrays offer broader spectral capture but pose demosaicking challenges.

Purpose of the Study:

• To develop an advanced demosaicking method for RGBW imaging systems.
• To improve low-light color imaging performance in applications such as night vision and surveillance.

Main Methods:

• Proposed a demosaicking method utilizing residual interpolation prior and a dual-branch decoding network.
• Developed a preprocessing interpolation algorithm to frame demosaicking as an image restoration task.
• Leveraged the W channel's sensitivity within a dual-branch network for optimized reconstruction.

Main Results:

• Significantly improved RGBW demosaicking performance under low-light conditions.
• Achieved reconstructed images with enhanced details and superior color fidelity.
• Demonstrated effectiveness through experiments using a custom low-light acquisition system and dataset.

Conclusions:

• The proposed method effectively addresses RGBW demosaicking challenges in low-light scenarios.
• The technique offers enhanced image quality closely matching human visual perception.
• This advancement is crucial for next-generation low-light imaging technologies.