Lossless compression of color sequences using optimal linear prediction theory.

Area of Science:

• Digital image processing
• Video compression algorithms
• Information theory

Background:

• Lossless compression is crucial for preserving video data integrity.
• Existing video compression techniques often struggle to fully exploit spatio-temporal and spectral redundancies.
• Optimal linear prediction offers a theoretical framework for efficient data modeling.

Purpose of the Study:

• To develop a novel lossless compression technique for color video sequences.
• To leverage optimal linear prediction theory to exploit all redundancies in video data.
• To improve prediction accuracy and minimize prediction error energy.

Main Methods:

• Utilizing optimal linear prediction theory for video compression.
• Estimating autocorrelation matrices incorporating spatial, spectral, and temporal correlations.
• Calculating cross-correlations between adjacent frames and color components.
• Coding the residual image using a context-based Golomb-Rice coder with quantized local prediction error variance for error modeling.

Main Results:

• The proposed algorithm achieves significant compression ratios for color video sequences.
• The method effectively exploits spatial, spectral, and temporal redundancies.
• The compression technique demonstrates robustness against scene changes.
• Reduced prediction error energy was observed due to improved prediction.

Conclusions:

• The novel technique offers an effective approach to lossless color video compression.
• Exploiting multiple correlation types enhances prediction accuracy and compression efficiency.
• The algorithm's robustness makes it suitable for real-world video compression applications.