Modality And Region-Oriented Medical Image Compression Using Wavelets
Medical image compression is becoming increasingly important with the emergence of telemedicine applications. The need for data compression is crucial for storage and transmission purposes since the medical images are inherently voluminous, much more bulky than the usual images sent over the interne...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Published: |
2003
|
| Subjects: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Medical image compression is becoming increasingly important with the emergence of telemedicine applications. The need for data compression is crucial for storage and transmission purposes since the medical images are inherently voluminous, much more bulky than the usual images sent over the internet. One of the popular compression techniques to use wavelets. Wavelets complement the shortcomings of JPEG and are better matched to the human visual system (HVS) characteristics.
Currently, quite a number of wavelet filters for images compression are being suggested by different authors, with different results for different images at different compression ratios. In medical images, different modalities have different image characteristics (noise, texture, intensity profile, sharpness, etc). Hence, in this thesis, we tested thirteen different wavelet filters for a large number of images(383 images) from different modalities (CT, MRI, X-ray and ultrasound) at eight different compression ratios. The objectives are to recommend one particular optimal wavelet filter per modality and per compression ratio and or to recommend one optimal wavelet filter for all medical images. This analysis is termed "modality-oriented medical image compression".
A medical image consists of different regions such as backgrounds, regions of interest (ROI)and contextual information. Each of these regions has different texture, dynamic range and noise. In the second phase, we tested different wavelets for different regions in a medical image, with the aim to find an optimal wavelets per region. This second analysis is termed "region-oriented medical image compression".
Observations showed that it is possible to recommend a particular wavelet filter for each imaging modality at each range of compression ratios. However, the use of specific filters per modality and compression ratio gives only marginal benefit compared to using one optimal wavelet, namely the Antonini ( or Daubechies 9, 7 wavelet) for all medical images, except for ultrasound images. Hence, the Antonini filter can be recommended as the optimal wavelet filter for all medical images, irrespective of modality and compression ratio.
In the case of region-oriented medical image compression, no pattern could be observed in the choice of optimum wavelet. This indicates that it is impossible to select a wavelet on the basis of specific image characteristics. Furthermore, the global performance decreases by choosing the best wavelet for each region compared to using one wavelet filter for the whole image. This apparently due to the classical border phenomena at the border of the tiles. Hence, the recommendation to avoid region-oriented compression when compression performance is the main criterion. |
|---|