Performance analysis of reversible image compression techniques for high-resolution digital teleradiology

Area of Science:

• Medical Imaging
• Image Compression
• Digital Signal Processing

Background:

• Very-large-format images (≥4096x4096 pixels) require efficient compression techniques.
• Existing algorithms like Huffman, Lempel-Ziv, and arithmetic coding have limitations for medical image data.
• Lossless compression is crucial for diagnostic accuracy in medical imaging.

Purpose of the Study:

• To compare the performance of various block-based, reversible compression algorithms.
• To evaluate a novel two-dimensional linear predictive coder for large medical image compression.
• To assess compression ratios and information preservation for mammograms and chest radiographs.

Main Methods:

• Implementation and comparison of Huffman coding, Lempel-Ziv coding, arithmetic coding, and transform coding (DFT, DCT, DWT).
• Development and testing of a novel two-dimensional linear predictive coder based on the multichannel Burg algorithm.
• Digitization of mammograms and chest radiographs to sizes up to 4096x4096 pixels (10 bits/pixel).

Main Results:

• Achieved lossless compression from 10 bits/pixel down to 2.5-3.0 bits/pixel for the tested medical images.
• The novel two-dimensional linear predictive coder demonstrated superior performance compared to other evaluated methods.
• The proposed method offers practical implementation advantages.

Conclusions:

• The novel two-dimensional linear predictive coder is highly effective for lossless compression of very-large-format medical images.
• This technique offers a significant improvement over existing compression algorithms for mammograms and chest radiographs.
• The method provides a balance of high compression ratios and implementation feasibility.