Lambrou, Tryphon; (2001) Wavelet transform in applications in digital mammography. Doctoral thesis (Ph.D), UCL (University College London).

Text Wavelet_transform_in_applicati.pdf Download (8MB)

Abstract

This thesis presents an investigation on the performance of all the 360 real 4-tap wavelet filters when used for mammographic image analysis and classification. The images are analysed by using the following Wavelet Transform architectures: Logarithmic Splitting, Wavelet Packet, and the Adaptive Splitting. The pattern recognition approach incorporates twenty statistical measurements, and five different classifiers. In addition the same features were collected from the time domain. Overall, 276 digitised mammographic extracted subimages (256x256x8), 138 abnormal and 138 normal, obtained from the Mammographic Image Analysis Society database were used in order to evaluate the performance of the wavelet transform analysis and classification. Our results suggest that: Firstly, the mean normalised variance values of the wavelet transform coefficients follow a specific pattern, which presents some differences among the normal and the abnormal images. This finding can be used for the selection of the appropriate wavelet filter for mammographic image analysis. Secondly, the adaptive wavelet transform method for mammographic classification provides higher classification accuracy than the other wavelet based techniques. Detailed comparison of the classification performance of the statistical features and the classifiers across the different domains is also provided.

Download activity - last month

Export as JSONCSVXML

Download activity - last 12 months

Export as JSONXMLCSV

Downloads by country - last 12 months

Export as JSONXMLCSV

Archive Staff Only

View Item