A Portable Capturing System for Image-based Relighting

A Portable Capturing System for Image-based Relighting
Wai-Man Pang

In this thesis, we present a novel capturing system for acquiring image-based object or scene from the real world. More specifically, we tailor the system for image-based relighting (IBL). Image-based relighting is a relatively new topic in the area of image-based modeling and rendering (IBMR). It allows the modification of lighting condition of an image-based scene. To prepare the image-based scene, one can render several images for the case of synthetic scene. However for the case of real scene, the data acquisition is difficult because the viewing and the light vectors have to be determined as well as the images.

While constructing the acquisition system, it is difficult to detect the lighting direction. Previous systems rely on mechanical framework for precisely controlling the lighting direction. However, they are difficult to setup, non-portable and high-cost. The setup also restricts the size of target objects. Instead, we employ the camera-based tracking technique to develop a self-calibrated, free-hand and portable acquisition system.

Our system consists of portable equipments including a light-weight web-cam, a hand-held spotlight, a checker board, a DV camera and an ordinary PC. By attaching the web-cam onto the spotlight, we can track the light vector by analyzing the orientation of checker board as seen from the web-cam. The captured illumination data is postprocessed and compressed, so that the image-based scene can be easily sent by email and relit on ordinary PCs.

We have used the capturing system to acquire different materials including furry object, anisotropic and glossy surfaces. Relighting are done on the captured image-based scenes to demonstrate the effectiveness of the proposed capturing system.