Lossless White Balance for Improved Lossless CFA Image and Video Compression

Area of Science:

• Image processing
• Computer vision
• Signal processing

Background:

• Color filter arrays (CFAs) are fundamental to modern color image sensors, enabling spatial multiplexing of color information.
• Current lossless compression techniques for raw sensor data rely on exploiting spatial and cross-color correlations, often using lifting schemes.
• White balance is a critical step in image processing, aiming to correct color casts.

Purpose of the Study:

• To propose a novel lifting-based lossless white balance algorithm.
• To investigate the effect of this algorithm on the spatial bandwidth of chrominance signals in raw sensor data.
• To evaluate its utility as a pre-processing step for enhancing lossless CFA subsampled image/video compression efficiency.

Main Methods:

• Development of a lifting-based algorithm specifically for lossless white balance.
• Application of the proposed algorithm to raw sensor data.
• Analysis of the impact on the spatial bandwidth of implied chrominance signals.
• Integration of the white balance algorithm as a pre-processing step for lossless CFA compression.

Main Results:

• The proposed lifting-based white balance algorithm effectively reduces the spatial bandwidth of chrominance signals.
• Applying this white balance pre-processing step demonstrably improves the coding efficiency of lossless CFA subsampled image/video compression.
• The algorithm maintains lossless compression of the raw sensor data.

Conclusions:

• The developed lifting-based lossless white balance algorithm is a valuable pre-processing technique for raw sensor data.
• This approach offers significant improvements in the overall coding efficiency of lossless CFA subsampled image and video compression.
• The method leverages the inherent properties of CFA data to enhance compression performance.