Yao-Jen Chang
Research Focus
-
Physics-based Computer Vision:
Reformulate the structure from motion framework to incorporate the refractive distortion model for multi-view 3D scene reconstruction under the refractive plane. -
Active Learning for Interactive 3D Reconstruction:
Apply the active learning framework widely adopted in CBIR to the interactive 3D scene reconstruction. By quantifying the uncertainty of each region’s geometry inferred by the energy minimization, our algorithm requests feedbacks from a user to provide intuitive geometric cues on the most informative regions; thereby effectively reducing human's efforts for interactive 3D reconstruction. -
Active Image-based Rendering with Robot-based Imaging Testbed:
Design mobile robots for active capturing and rendering of environments. -
Embedded Video Surveillance:
Investigate distributed video surveillance networks based on miniaturized smart cameras capable of performing on-board visual analysis and video streaming. -
Software Development Toolkit for Video Surveillance:
Investigate a video surveillance architecture and framework for rapid development of video surveillance applications for home care, home security, and parking lot services. -
Image-based Videorealistic Speech Animation:
Investigate a user adaptation method for creation of image-based speech animation of a novel person with limited video corpus. -
Algorithms for Key Generation based on Biometrics:
Develop a biometrics-based cryptographic key generation mechanism for generating cryptographic keys on resource-constrained handheld devices.