Image Analysis and Vision Research Old Well

The University of North Carolina at Chapel Hill
College of Arts and Sciences
Department of Computer Science
Medical Image Display & Analysis Group

Image Analysis and Vision Research at UNC

Home page of the Medical Image Display and Analysis Group

Introduction to research in image analysis and vision

People, Courses, and Key Ideas in image analysis and vision

Projects in Image Analysis and Vision

Mixture Modeling for Medical Image Segmentation

In mixture modeling, multiple "component" distributions are combined to represent complex distributions of data. We are particularly interested in the definition of mixture models comprised of an infinite number of Gaussian components whose collective parameterizations are continuous. We have found these models to be ideal for representing the data of medical images and color spaces.

Object & Shape Based Medical Image Analysis

Objects need to be extracted from medical images for visualization and measurement. Images need to be geometrically transformed to bring the same object in each image into registration. The shape of an object needs to be measured to allow diagnosis and to focus extraction on an object of the desired shape. We focus on models of objects and object shape involving medial, i.e., solid figural, aspects, but also including boundary and landmark aspects. Much of our work involves deformable models, but direct analysis of images via medial atoms and height ridges of medial strength are also featured.

Statistical Features for Multiscale, Geometric Image Analysis (Spatial Spectroscopy)

In Spatial Spectroscopy, measurements of local image structure (intensity or derivatives) at multiple scales are used as features (in the sense of statistical pattern recognition) to describe the image. Inferences about image content are based on these features.
This page is maintained by the Department of Computer Science.
Send comments to Dr. Graham Gash
Last update: 16 October 1997