Advisor: Layla Oesper
As humans, we’re really good at looking at an image and being able to identify and distinguish different parts of the picture. At its most basic level, this process is called image segmentation. This is where we take an image and assign a label to every pixel in the image so that pixels with the same label share some characteristic (color, texture, etc.).
There are a number of important domains where automated image segmentation is essential. For example, in the medical realm, image segmentation can be used to help identify tumors from an MRI or CT scan. A more precise understanding of the location and boundaries of a tumor (as well as other features such as whether or not blood vessels are feeding the tumor) can be helpful to physicians as they determine a plan of treatment for a patient. Other applications of image segmentation include object detection, recognition tasks such as face or fingerprint recognition, certain aspects of traffic control, or even things as mundane as helping a homeowner visualize what their dining room will look like when painted blue (as I recently discovered!).
(Left) An original image of tennis players and a segmented version. (Right) A chest x-ray of a patient with pneumonia and a segmented version of the lungs in the image.
In this project you will explore algorithms for and applications of image segmentation. In particular you will:
Courses that may be useful for this project include algorithms, linear algebra, AI, data mining, graphics, and computational models of cognition.
Below are a few papers about existing work in image segmentation. These are only intended to provide you a minimal start for your literature search - they are certainly not the only nor necessarily the best sources for ideas. You will be finding and reading additional sources!
Tuesday/Thursday 2:45pm - 3:45pm for Fall and Winter