Compressing illumination adjustable images

Abstract

In an image-based re-lighting (IBL) system, we can control the lighting condition of a scene. A new image under a novel lighting condition is synthesised from a number of pre-recorded reference images captured under known lighting conditions. In the system, synthesising a new image is independent of the scene complexity. A tremendous storage of the reference images should be required to make high-quality of images. The collection of the reference images is renamed as IBL data set. Effective and efficient compression techniques applying on the IBL data can make a faster and more practical interactive or real time relighting engine in the rendering situation. Within the IBL data set, there is a close relationship among each reference image. Based on the property of the IBL data set, we propose using an eigen-based method to exploit the correlation of the data. A wavelet-based approach will be applied for further compression.