Scalar-vector quantization of medical images

Area of Science:

• Medical Imaging
• Image Compression
• Digital Radiology

Background:

• Traditional image compression methods can introduce complexity and error propagation, especially over noisy channels.
• Picture Archiving and Communication Systems (PACS) require reliable, high-fidelity image transmission and storage.
• Existing entropy-constrained scalar quantizers (ECSQs) offer good performance but can be complex.

Purpose of the Study:

• To develop a new, efficient coding scheme for medical image compression using scalar-vector quantization (SVQ).
• To evaluate the performance of the SVQ-based scheme against optimal entropy-constrained scalar quantizers (ECSQs).
• To assess the suitability of the SVQ scheme for Picture Archiving and Communication Systems (PACS) in digital radiology.

Main Methods:

• Developed a novel coding scheme based on the scalar-vector quantizer (SVQ).
• Compared the rate-distortion performance of SVQ with ECSQ for memoryless sources.
• Tested the coding scheme on a set of magnetic resonance (MR) images.
• Evaluated image quality perceptually on a monitor.

Main Results:

• The SVQ scheme demonstrates rate-distortion performance comparable to optimal ECSQs.
• At low bit rates, SVQ and ECSQ coding results for MR images are indistinguishable.
• Encoded images are perceptually indistinguishable from the original when displayed.
• The SVQ coder does not suffer from the error propagation typical of variable-length codes over noisy channels.

Conclusions:

• The SVQ-based coding scheme is an attractive option for medical image compression.
• Its fixed-rate nature simplifies implementation compared to variable-length codes.
• The scheme offers high fidelity and robustness, making it suitable for PACS in all-digital radiology environments.