Algorithms For Compressing Ultrasonic ImagesAnd Audio Doppler Ultrasound At Very Low BitRates

Abstract

This paper presents and evaluates a variety of algorithms for compressing ultrasonic images and baseband audio doppler ultrasound to very low bitrates. In the case of monochrome ultrasonic image compression a number of single-frame and inter-frame codec algorithms are described and evaluated. For single-frame compression, schemes based on Differential Pulse CodeModulation (DPCM) and Entropy Encoding as well as the Discrete CosineTransform (DCT) and zonal coding are dealt with. The best compression figures that have been achieved on cardiac images are 2.5 bits/pel forthe DPCM/Entropy Coding scheme and 0.8 bits/pel for the DCT/Zonal Coding method. A range of inter-frame codec algorithms have been investigated based on Conditional Replenishment and hybrid DPCM/DCT schemes with full and partial motion compensation. The compression figures that have been achieved whilst still maintaining acceptable image quality range from 0.9bits/pel down to 0.22 bits/pel and the highest compression schemes allowthe transmission of motion cardiac signals at a rate of 1 6 frames/second over Basic Rate ISDN (128 Kbits/s). The paper also shows how the technique of linear prediction can be effectively applied to the compression of audio doppler ultrasound. The doppler signal is modelled as the output of a time-varying linear predictive filter excited by a random noise source. Both scalar and vector quantisation of the predictor coefficients have been investigated and it is shown that perceptually lossless compression of the audio doppler signal can be achieved down to bit rates of 1600 bits/s in thecase of scalar quantisation and 800 bits/s in the case of vector quantisation. Transactions on Biomedicine and Health vol 3, © 1996 WIT Press, www.witpress.com, ISSN 1743-3525