Space-frequency quantiser design for ultrasound image compression based on minimum description length criterion.

Area of Science:

• Medical Imaging
• Signal Processing
• Computer Vision

Background:

• Ultrasound (US) image compression is crucial for efficient storage and transmission.
• Existing methods struggle to optimally balance compression rates and image quality.
• Wavelet coefficient statistics in US images are complex and require specialized modeling.

Purpose of the Study:

• To jointly optimize spatial quantization and subband adaptive uniform scalar quantization for ultrasound images.
• To develop a novel compression framework based on the Minimum Description Length (MDL) principle.
• To improve the rate-distortion performance of ultrasound image compression.

Main Methods:

• Utilized the generalized Student's t-distribution to model wavelet coefficients in US images.
• Developed a Space-Frequency Quantizer (SFQ) named MDL-SFQ by integrating statistics with the rate-distortion (RD) criterion.
• Employed efficient zero-tree quantization for zero coefficients and adaptive scalar quantization for non-zero coefficients.

Main Results:

• The MDL-SFQ algorithm achieved variable bit-rates, supporting near-lossless to lossy compression.
• Experimental results demonstrated superior quantitative and qualitative performance compared to the SPIHT coder.
• Achieved an average improvement of 1.01 dB over the SPIHT coder at 0.25 bits per pixel.

Conclusions:

• The proposed MDL-SFQ framework offers enhanced compression efficiency for ultrasound images.
• Joint optimization of quantization parameters within the MDL framework is effective.
• The generalized Student's t-distribution provides a better statistical model for US image wavelet coefficients.