الفهرس 
Introduction
ObjectivesOnly 14 pages are availabe for public view
 |  | from | 210 |  |  |
| | | from | 210 | | |
Abstract
The study investigates and presents wavelet based compression schemes for mammograms. The mammograms are preprocessed by using a global segmentation technique for mammograms which segments the tissue area from the background area. The segmentation technique is based on modeling the background and subtracting it from the whole mammogram. The used segmentation technique yielded very good segmentation results of 90% for both accurate detection of a mammogram border and small excessive parts. The performance of three techniques based on the wavelet transform were investigated; Cohen-Daubechies-Feauveau (CDF) integer-to-integer wavelet transforms, quincunx wavelets using the Neville filters, and the adapted wavelet transform using locally adapted estimation method (LAE). Moreover, two wavelet coefficients encoding schemes were considered; Set Partitioning in Hierarchical Tress (SPIHT) and the Region-Of-Interest SPIHT (ROI-SPIHT). SPIHT was used to encode the whole image with progressive encoding capability, whereas the ROI-SPIHT is also used to encode the whole image but with higher fidelity of the tissue area. The performance of the investigated wavelet based compression schemes is compared to the state-of-art compression schemes; CALIC and LOCO-I. Sixty mammograms from the MiniMIAS database are used for judging the performance of the investigated schemes. Results show that both the LAE and quincunx wavelet transforms encoded with the SPIHT give very close performance to the state-of-art encoding schemes, within 3.52 % and 9.14 % from the CALIC and LOCO-I respectively, and with the advantages of progressive and ROI encoding capabilities. Moreover, it has been found that the wavelet based compression schemes, LAE and quincunx, outperform the CALIC compression scheme in 15 % of the tested images. Such results make the wavelet based compression schemes proper candidates for long-term archiving purposes. Moreover, the quality of the results obtained by encoding only the mammogram tissue area using the ROI-SPIHT has shown the importance of ROI encoding of mammograms, specially for teleradiology applications. The obtained results indicate that the two-dimensional and adaptive wavelet transforms combined with subband encoding schemes such as SPIHT and ROI-SPIHT form a promising trend for developing Picture Archiving and Communications Systems (PACS) for mammograms. Key Words: Medical Image Compression – Wavelet Image Compression – Mammogram Compression