Lossless Image Coding Using Non-MMSE Algorithms to Calculate Linear Prediction Coefficients

Area of Science:

• Computer Science
• Image Processing
• Data Compression

Background:

• Existing lossless image compression methods often face trade-offs between compression efficiency, decoding speed, and implementation complexity.
• Optimizing prediction coefficients and prediction error coding are key areas for improving codec performance.

Purpose of the Study:

• To develop a lossless image compression method with significantly faster decoding times and flexible parameter adjustment.
• To compare different approaches for computing non-MMSE prediction coefficients.
• To enhance data modeling and prediction error coding stages for superior compression performance.

Main Methods:

• The data modeling stage employed both linear (non-MMSE) and non-linear predictions, including a context-dependent constant component removal block.
• Prediction error coding utilized a two-stage compression approach: adaptive Golomb coding followed by binary arithmetic coding.
• A comparative analysis of various non-MMSE prediction coefficient computation methods was conducted.

Main Results:

• The proposed method achieved 30% shorter decoding times compared to competing solutions.
• The codec demonstrated a 7.9% lower average bit rate relative to the JPEG-LS codec.
• Flexible adjustment of coder parameters allowed for control over implementation complexity.

Conclusions:

• The developed lossless image compression method offers a compelling balance of speed, compression efficiency, and flexibility.
• The optimized prediction and error coding strategies contribute to significant performance gains.
• This approach presents a viable alternative for applications requiring fast and efficient lossless image compression.