Wavelet- versus JPEG Compression of Echocardiograms
C.H. Jansen, L. Zeng, M. Arigovindan, M. Sühling, S. Marsch, P.R. Hunziker
Proceedings of the Fifty-Second Annual Scientific Session of the American College of Cardiology (ACC'03), Chicago IL, USA, March 30-April 2, 2003, [Journal of the American College of Cardiology, vol. 41, no. 6, supp. 2, March 19, 2003], pp. 449A.
Background: Image compression of echocardiograms is important because it allows fast data transmission across slow lines in telemedicine and limits the expense for storage media. Although wavelet compression becomes increasingly important, there is limited data about its value in echo, whereas JPEG compression is well established.
Methods: Nonselected clinical echos were subjected to JPEG compression, to wavelet compression (using biorthogonal wavelet filters and a zerotree compression strategy) and compared to noncompressed loops. Compression quality was judged quantitatively by comparing the signal-to-noise ratio of compressed vs noncompressed data, as well as qualitatively by expert review of the loops(Visual scale 10 best, 0 worst). Compression rates from 20:1, 32:1, 64:1, 128:1 and 300:1 were tested.
Results: Compression results are given in Table. Already at low compression rate, quantitative error analysis shows an advantage of wavelet compression, although the visible loss is small (apart from JPEG block artifacts) with both methods. At higher compression rate, image quality deteriorates less rapidly with wavelets than with JPEG. Very high compression (128-300:1) is feasible only with wavelets.
Conclusion: Compared to JPEG, wavelet compression of clinical echo loops yields better compression quality, especially at higher compression rates. For telemedicine across connections with limited bandwidth and in cost-sensitive storage applications, wavelets should be preferred.
| JPEG vs Wavelets: Subjective and Objective Results | |||||
|---|---|---|---|---|---|
| Compression strategy/rate | JPEG visual | Wavelet visual | p | Wavelet SNR | JPEG SNR |
| 20:1 | 8.8±0.90 | 8.9±0.7 | 0.68, n.s. | ||
| 32:1 | 8.8±0.80 | 8.5±0.7 | 0.44, n.s. | 37.69 | 21.88 |
| 64:1 | 6.0±1.25 | 7.2±0.6 | <0.001 | 36.41 | 13.17 |
| 128:1 | 3.5±1.13 | 6.5±1.1 | <0.001 | 34.62 | 11.71 |
| 300:1 | 1.33±0.50 | 5.1±1.0 | <0.001 | 33.73 | 9.84 |
@INPROCEEDINGS(http://bigwww.epfl.ch/publications/jansen0303.html,
AUTHOR="Jansen, C.H. and Zeng, L. and Arigovindan, M. and S{\"{u}}hling,
M. and Marsch, S. and Hunziker, P.R.",
TITLE="Wavelet- {\textit{versus}} {JPEG-} Compression of
Echocardiograms",
BOOKTITLE="Fifty-Second Annual Scientific Session of the American
College of Cardiology ({ACC'03})",
YEAR="2003",
editor="",
volume="41, no. 6, supp. 2, March 19, 2003",
series="Journal of the American College of Cardiology",
pages="449A",
address="Chicago IL, USA",
month="March 30-April 2,",
organization="",
publisher="",
note="")