Loading
Research Projects
Segmentation of Biological Microscopy Images
|
Abstract
We will address the problem of segmenting shapes from biological images and will focus on images of fruit fly denticles and yeast cells as motivating examples. An accurate segmentation of these images helps biologists make calculations and draw scientific conclusions. Issues in the segmentation of images in biology arise due to varying position, size, contrast and shape of the objects that need to be segmented. Segmentation can also be made more difficult by the appearance of adjacent objects. We propose to combine a global, triangulation based approach by Felzenszwalb with Blake and Isard's local search, active contour approach. The triangulation based approach yields locations of all the shapes in the image, while the active contour approach will be used to refine the solution by obtaining a better fitting closed curve around every detected shape.
Advisor(s): Carlo Tomasi
Pankaj Agarwal, Alex Hartemink