Research projects - from the most recent
The goal for this project is to quantify the number of "things" in
an image, which we call the proto-objects - image regions that
have consistent low-level features. Proto-objects are segmented
using a novel parametric graph partitioning method, and used
to model human visual clutter perception. The method can also
be used as a clustering algorithm.
The precise segmentation of brain tumors from MR images is necessary
for surgical planning. However, it is a tedious task for the medical professionals
to process manually. This project is an unsupervised, fully automatic brain
tumor detection and segmentation method based on utilizing the blob-like
characteristics, as well as the brain billateral asymmetry that results from the
presence of tumors.
The neurons of the Medial Superior Temporal (MST) area are
parts of the "where" pathway of the dorsal stream, in visual
processing system. They are responsible for the perception of
self-motion, and we explore ways to model the MST neuron
receptive fields using a dual-Gaussian models.
This is the final project for the Advanced Computer Vision course
at Penn State University, Spring 2009, taught by Dr. Robert Colllins.
We focused on object tracking using deformable contours, and we
implemented the Hidden Markov Model method by Huang et. al 2006,
and tested on hand movements as well as brain MRI frames.
A virtual-reality of the Strong Memorial Hospital (URMC) at Rochester, NY.
It is a system with various different tasks-oriented tests for assessing a
person's navigation impairment. The virtual-reality was constructed using
Quake 3 game engine in C, and all the tests were implemented using the