Low Cost Edge Sensing for High Quality Demosaicking

Area of Science:

• Digital imaging and computer vision.
• Image processing algorithms.
• Computational photography.

Background:

• Digital cameras utilize Color Filter Arrays (CFA) requiring demosaicking to generate full RGB images.
• Demosaicking speed and accuracy are critical for user experience and cost-effectiveness in the digital camera industry.
• The classical Hamilton-Adams (HA) algorithm offers efficiency but limited accuracy due to simplified smoothness formulations.

Purpose of the Study:

• To improve the accuracy and speed of demosaicking algorithms for CFA images.
• To develop a cost-effective demosaicking solution that balances performance and computational cost.
• To enhance the Hamilton-Adams algorithm by incorporating an edge-sensing mechanism.

Main Methods:

• A novel, low-cost edge sensing scheme was developed, guiding demosaicking using a logistic function of directional variation differences.
• The proposed method was extensively compared against 27 existing demosaicking algorithms using open-source implementations on benchmark datasets.
• Performance was evaluated based on demosaicking accuracy and computational cost.

Main Results:

• The proposed algorithm achieves substantially higher accuracy compared to methods with similar computational costs.
• Compared to algorithms with similar accuracy, the new method offers significantly lower computational cost.
• For popular image resolutions, the algorithm's quality is comparable to top-performing methods, but with tens of times faster processing speed.

Conclusions:

• The developed edge sensing scheme effectively enhances demosaicking accuracy without a significant increase in computational cost.
• This approach provides a superior trade-off between speed, accuracy, and cost for CFA demosaicking in digital cameras.
• The findings suggest a practical advancement for real-time image processing in digital imaging applications.