On exploiting geometric constraint of image wavelet coefficients.

Area of Science:

• Digital image processing
• Signal processing
• Computer vision

Background:

• Wavelet-based image coding is crucial for efficient data compression.
• Existing methods often use zero-mean models for wavelet coefficients, limiting coding efficiency.
• Exploiting geometric constraints, particularly edge information, offers potential for improved modeling.

Purpose of the Study:

• To investigate the exploitation of geometric edge constraints in wavelet-based image coding.
• To develop novel algorithms for phase shifting and prediction in the wavelet domain.
• To improve the probabilistic modeling of high-frequency wavelet coefficients for enhanced coding gain.

Main Methods:

• Derivation of novel phase shifting and prediction algorithms within the wavelet space.
• Development of biased-mean probability models for high-band wavelet coefficients.
• Quantitative analysis of coding gain within a differential pulse-code modulation (DPCM) framework for lossy coding.

Main Results:

• Phase uncertainty in wavelet coefficients can be resolved, enabling better modeling.
• Biased-mean probability models significantly outperform existing zero-mean models for high-band coefficients.
• The proposed phase shifting and prediction scheme demonstrably improves both subjective and objective performance of wavelet-based image coders.

Conclusions:

• Exploiting geometric edge constraints enhances wavelet-based image coding.
• Biased-mean models provide a more accurate representation of wavelet coefficients, leading to coding gains.
• The developed algorithms offer a practical improvement for image compression systems.