About AMP Lab Projects Downloads Publications People Links
HyunJung Shim Personal
Homepage: http://amp.ece.cmu.edu/people/Kate/ |
||
Office:
Porter Hall B26 Lab: Porter Hall B6 Phone: 412-268-7100 Fax: 412-268-3890 |
Mailing
Address: Department of ECE, Carnegie Mellon University, 5000 Forbes Avenue, Pittsburgh, PA 15213-3890 |
[Research Interests] [Project] [Publications]
Research Focus:
We develop the statistical framework for Image-Based Relighting. Using image-based relighting, one can render realistic relit images of a scene without prior knowledge of objects in the scene. However, traditional image-based relighting methods require too many input images, each corresponding to a lighting pattern. From the statistical analysis, we are able to build an efficient relighting system by spending least input images with the minimum mean square error.
Multi-view images have been important resources to many applications in computer vision. In this project, we present a multi-view capturing system that uses single camera but multiple mirrors. We place multiple mirrors at arbitrary positions and capture an image of those mirrors. From the image of multiple mirrors, we are able to extract multi-view images of a scene reflected by mirrors. The contribution of this work is to introduce a light weight implementation for multi-view capturing system.
Develop the Image-Based Relighting
System
We use a CRT monitor as a emitter of light source and a static digital camera (Canon G5) as a receiver of illumination about the scene object. Captured images from shown set-up are to get the information about the scene properties. Our image-based relighting system captures images automatically with entire control of camera. By using of statistically derived lighting patterns, we can achieve highly realistic relit images with the least number of training images.
Please refer to the Image-Based Relighting System project page for more details. |
Light Weight Multi-View Capturing We present a light weight multi-view capturing system with single camera and 25 planar mirrors as shown in the left figure. By capturing single image which contains 25 images, we could capture multi-view images, each corresponds to a mirror.
Please refer to the Mirror array project page for more details. |
Journal Papers:
Conference Papers:
This website
is maintained by Devi Parikh
|