Research Interests:

• Computer Vision, Object Segmentation and Recognition
• Machine learning
• Medical Image Segmentation and Registration
• Content Based Image and Video Retrieval
• Multimedia Database System

Projects Involved:

• Robust Image Semgntation: I am interested in all types of image segmentation algorithms: from region based methods, such as mean-shift and graph-cut, to contour based methods such as deformable models. I am paticularly interested in object based segmentation or so called object-cut. I recently developed a novel algorithm for robust multiple class object segmentation using Mean-shift patches and spatial bag of visual words model.
• Generic Object Recognition: A challenging problem in computer vision which I am quite interested.
• Machine Learning in Medical Image Processing: I am interested in applying modern machine learning algorithm to analyze medical images. We have already applied some algorithms and build computer aided diagonosis systems for early detection and prediction of blood and breast cancers.
• Content Based Image Retrieval - Pathminer Project: I developed a new version of the IGDS system for the Pathminer project. The software, which is written with Java, allow the users to control a robotic microscope remotely. Search engines for the database allow one to automatically identify and retrieve images, diagnoses and correlated clinical data of cases from a "gold standard" database, which exhibit spectral and spatial profiles most similar to the given query image. The system suggests the most likely diagnosis based on majority logic of the retrieved cases. The details of the new IGDS system and segmentation algorithm can be found in my IEEE Trans. 2005 paper.
• IBM - Help Defeat Cancer: The Help Defeat Cancer project comprises a large-scale effort between CBII and researchers at IBM to analyze features in cancer tissue microarrays. The goal of this research is to discover a unique signature which characterizes cancer in the digitized pathological images.
• Medical Image Segmentation and Registration: Previously I was working on the medical image segmentation using deformable models. I was also working on mutual information based image registration and segmention guided registration of CT and MRI images.
• Dynamic Behavior Analysis and Scheduling of Real-time Systems: Previously I was working on the resource allocation and performace prediction in the real-time system. I was also working on load-balancing. We developed an intelligent resource manager and load balancing system based on a distributed and prediction model using robust estimation.