[Symbol: see text]2 Optimized predictive image coding with [Symbol: see text]∞ bound.

Area of Science:

• Digital image processing
• Image compression algorithms
• Rate-distortion theory

Background:

• Existing image compression methods struggle to balance L-infinity (worst-case) error bounds with L2 (average) fidelity.
• L-infinity optimization can lead to structured errors and poor rate control, while L2 optimization lacks individual error bounds.

Purpose of the Study:

• To develop a new image compression approach that combines the benefits of L-infinity and L2 error metrics.
• To overcome the limitations of pure L-infinity and L2 optimized image coding methods.
• To improve rate control and reduce artifacts in near-lossless image compression.

Main Methods:

• Replaced uniform scalar quantizers in DPCM prediction loops with context-based L2-optimized quantizers.
• Optimized quantizers to minimize a weighted sum of L2 distortion and entropy.
• Ensured a strict L-infinity error bound is maintained throughout the compression process.

Main Results:

• The proposed method achieves good rate-distortion performance for both L2 and L-infinity metrics.
• It offers improved rate granularity compared to traditional methods.
• Guarantees lower L-infinity error and higher Peak Signal-to-Noise Ratio (PSNR) than JPEG 2000 at high bit rates.

Conclusions:

• The new compression approach effectively integrates L2 and L-infinity error optimization.
• It provides a superior alternative to existing methods like JPEG 2000 for applications requiring strict error bounds and high fidelity.
• The method enhances image compression quality and control for scientific, medical, and defense applications.