Faculty profile


Charles Dyer

Charles Dyer

Dept. of Computer Sciences
6379 Computer Sciences and Statistics
(608) 262-1965

dyer@cs.wisc.edu
Links: Department, Lab

Research Keywords

computer vision, image analysis, image processing

Affiliations

  • Department of Computer Sciences, Professor
  • Department of Biostatistics and Medical Informatics, affiliate
  • Eye Research Institute, leadership committee member
  • Computation and Informatics in Biology and Medicine Training Program, trainer

Current Projects

Research Collaborators

  • Jerry Zhu, Computer Science
  • Nigel Boston, Mathematics
  • Yu-Hen Hu, Electrical and Computer Engineering
  • Andrew Alexander, Medical Physics
  • Terrence Oakes, Waisman Laboratory for Brain Imaging and Behavior
  • Michael Gleicher, Computer Sciences
  • Li Zhang, Computer Sciences

Representative Classes

Research Statement

My research area is computer vision, i.e., developing computer algorithms for automatically analyzing images to extract descriptions of what is in the scene. Current projects include:

Face recognition and multi-camera networks. This is a collaborative project with Profs. Nigel Boston (Math) and Yu-Hen Hu (ECE). We have developed a new representation of 3D objects that combines images taken from multiple viewpoints. For example, a person rotates their head in front of a video camera and the center columns of pixels from each image frame are concatenated to form a "cyclograph" image in which each part of the face is taken from an approximately fronto-parallel viewpoint. Given a new image or video of an unknown face, we have developed algorithms for matching the input with a set of cyclograph images of different people in order to perform face recognition. Other work in the group has defined a new set of summation invariant features that can be used to recognize faces irrespective of pose.

Text-to-picture synthesis. The goal of the research with Prof. Jerry Zhu (CS) is to develop general-purpose algorithms that automatically generate pictures from natural language sentences, so that the picture conveys the main meaning of the text. The process might generate static or animated pictures that contain important objects referenced in the text. Additionally, it will represent spatial relations and actions that are implied in the text. Unlike prior systems that require hand-crafted narrative descriptions of a scene, our aim is to convert general, unrestricted text into understandable pictures. The central hypothesis of this project is that general text-to-picture (TTP) conversion can be achieved to the extent that it will be useful for multiple applications, ranging from improving children's reading comprehension to helping people with communication disabilities. Our approach uses statistical machine learning, and draws ideas from automatic machine translation, text summarization, text-to-speech synthesis, computer vision, and graphics.

Medical image analysis. In collaboration with Prof. Andy Alexander (Medical Physics) and Terry Oakes (Keck Lab for Brain Imaging and Behavior), we are investigating several problems involving the analysis of magnetic resonance images (MRI). We are studying the problem of extracting and reconstructing the 3D structure of white matter fiber tracts from a set of DTI images of the brain, which characterize anatomical pathways in the brain. We are also developing a novel structural model of the brain called a “nodetree,” which consists of a series of nodes that capture multiple scales of the brain’s structure from MRI images.

Selected Publications

  • A Text-to-Picture Synthesis System for Augmenting Communication. X. Zhu, A. B. Goldberg, M. Eldawy, C. R. Dyer, and B. Strock, Proc. 22nd Conf. on Artificial Intelligence: Integrated Intelligence Track, 2007.
  • Face Cyclographs for Recognition. G-D. Guo and C. R. Dyer, Proc. 8th Int. Conf. on Computer Vision, Pattern Recognition & Image Processing, 2007.
  • Learning from Examples in the Small Sample Case: Face Expression Recognition. G-D. Guo and C. R. Dyer, IEEE Trans. Systems, Man, and Cybernetics, Part B: Cybernetics 35(3), 2005, 479-488.
  • Linear Combination Representation for Outlier Detection in Motion Tracking. G-D. Guo, C. R. Dyer, and Z. Zhang, Proc. Computer Vision and Pattern Recognition Conf., Vol. 2, 2005, 274-281.