Digital image processing

Research staff

Staff

Collaborators

Abstract

The research attends to several issue inherent different aspects of the image processing, such as digital image noise estimation, filtering, object classification.
At present, the main study is applied to medical image processing, and, in particular, to denoising and enhancement of mammographic images by means of redundant wavelet transforms, the so called discrete dyadic wavelet transforms.
Mammography is the most effective method for early detection of breast diseases, however 10-30% of breast lesions are missed during routine screening. The most important signs of breast cancer are masses and microcalcifications. Small masses are difficult to detect because of the minor difference in x-ray attenuation between normal and malignant glandular tissue, while microcalcifications, although they have high inherent attenuation properties, appear with low contrast due to their small size.
In recent years, in order to increase radiologists diagnostic performance, a lot of computerized image analysis techniques have been developed with the aim of enhancing mammographic features, without emphasizing noise.
Among them, enhancement methods based on wavelet transform proved to be very useful because of their multi-resolution properties. In fact, in mammograms different features, such as microcalcifications, masses, background, and noise, appear at different scales, and so they can be selectively enhanced, detected or reduced within different resolution levels.
At present, the research is based on the DDSM (Digital Database for Screening Mammography) at the University of South Florida.

Publications

Theses