Abstract:
Medical image compression addresses the issues of larger storage and faster transmission requirements. Contextual compression of Ultrasound (US) medical images aims at compressing the diagnostically important region, region of interest (Rol), of an image with supreme quality as compared to rather unimportant area i.e. the background. Thus, the Rol area is compressed with less compression ratio and the background with the highest possible compression ratio in order to get better overall compression performance.
As a part of contextual compression technique four compression algorithms viz. JPEG coding, Wavelet Transform coding, JPEG 2000 coding and the SPIRT coding algorithms have been implemented in the present work. These algorithms have been implemented using MATLAB, Image Processing and Wavelet toolbox, in particular. A detailed analysis on the basis of parameters like mean square error (MSE), peak signal to noise ratio (PSNR) and correlation coefficient (CoC) has been carried out to evaluate these algorithms.
It is observed that JPEG 2000 and SPIHT compression algorithms perform better at lower bit rates (i.e. bpp below 0.20), whereas JPEG coding should not be used for lower bit rates. The Wavelet transform coding has outperformed the rest of the compression algorithms in terms of MSE, PSNR, CoC and of course, the visual quality of a compressed image. The Wavelet transform has an excellent energy retaining capability, which is responsible for its outstanding performance. In fact, the other two Wavelet based coding schemes viz. the JPEG 2000 and SPIHT too, are not far behind in performance.
The contextual compression scheme has achieved a compression ratio (CR) of as high as 65.24:1 an US image, whereas a simple or conventional compression scheme could hardly achieve the CR of 40:1. The prime contribution of a contextual compression scheme over a conventional compression scheme is improvement in CR and visual quality of an Rol of a compressed image with modest values of MSE and PSNR.
