Image coding with an Linfinity norm and confidence interval criteria

Area of Science:

• Digital image processing
• Information theory
• Computer vision

Background:

• Traditional image coding methods often face limitations in balancing compression efficiency and reconstruction fidelity.
• Existing techniques may not offer sufficient flexibility for both lossless and controlled lossy compression scenarios.

Purpose of the Study:

• To investigate a new image coding technique.
• To leverage the Linfinity-norm criterion and statistical properties of reconstruction error.
• To achieve flexible lossless and controlled lossy image compression.

Main Methods:

• Image preprocessing using linear prediction and iterated filterbanks.
• Quantization and encoding of image data.
• Reconstruction of images within a specified confidence interval.
• Exploitation of statistical properties of the reconstruction error.

Main Results:

• The proposed technique allows for both lossless and controlled lossy image coding.
• Reconstruction error is managed within a defined confidence interval.
• The method is compatible with existing image coding techniques.
• Controlled lossy compression can be achieved with a specified percentage of differing pixels.

Conclusions:

• The developed image coding technique offers significant flexibility.
• It effectively balances compression and fidelity through the Linfinity-norm criterion.
• The approach is adaptable and integrates with prior methods for enhanced performance.