Introduction

Aside from lossless compression techniques, lossy compression algorithms become more and more important in the field of image data communication.
One of the major disadvantages of lossless compression is its relatively low compression rate when being used on image data. With any lossless algorithm, average image data can be compressed with a ratio of 1:2.5-1:3 at maximum, whereas lossy compression yields considerably higher compression rates. However, the loss of information leads to artifacts in the compressed image.
Since in the field of radiology one has to deal with huge amounts of image data, a suitable compression algorithm should be fast and the loss of image information should remain as low as possible with increasing compression rates.

Lossy Compression

JPEG, fractal, and wavelet compression represent typical lossy compression techniques.
JPEG compression is rather fast, but already with relatively low compression rates the typical blocky artifacts of this type of compression become obvious in the processed image data.
With fractal compression, the computational power needed rises exponentially in relation to the compression rate, and the obtained data is very complex. Thus, this technique is not suitable for online use.
Wavelet compression achieves very high compression rates, the data does not become very complex. Furthermore, the processing time remains short; this is a very important factor with respect to any online (teleradiological) application.

Results Obtained Using Own CT and MRI Image Data.

By means of a standard wavelet algorithm that has been supplied by Summus, Ltd. and was implemented on a Picos-Workstation (Atec GmbH, Kassel) we have evaluated the impact of wavelet compression at various rates on image data, taking into consideration the overall impression and the diagnostic quality of the extracted data.
Typical CT and MRI images have been compressed at various rates. Ten radiologists have been asked to evaluate the data and aswer the following questions:
At which stage does the impact of the compression become obvious?
At which stage does the image lose its full diagnostic quality?
Our results yielded that the behavior of the wavelet compression is directly associated with the contents of the image, especially its contrast and its signal-to-noise ratio.
Typical low contrast images, e.g. CTs of the head or spine, tolerate lower compression rates between 1:6 and 1:10 at best, while high contrast images, e.g. CTs of the sinuses, the petrous bone, or the lung tissue, can be compressed with a ratio up to 1:50-1:60. The average compression rate of other clinical images is to be found between those extremes.

In order to illustrate the above we have set up a page that has some examples on it.

Summary

According to our first results the wavelet compression can be understood as a key technology in image archiving (keyword PACS) and image data communication, the latter especially with respect to a teleradiological point of view.
Using a sophisticated wavelet algorithm, image data compression minimizes the costs associated with archiving and online applications to a major degree; the network performance is improved considerably due to the small amount of image data transmission time needed.

[Team] [Services] [Equipment] [Projects]
[Informations] [Office Hours] [Appointments] [Site Plan]
[Home]