Disparity Estimation for Light Field Imaging Superimposed with Reflections

Depth image based rendering from 4-D light field data superimposed with reflections fromtransparent glasses requires the reconstructed depth information. In this case the existing Lambertainsurface assumption based disparity estimation schemes usually fail, and the related research topics suchas occlusion detection and occlusion handling are to be explored. Thus, the main goals of thisone-year project consist of three parts and they are stated as follows. (1) This project will propose arobust occlusion detection scheme for light field data with reflections from transparent glasses,especially for depth-discontinuity regions near the border of a slanted surface. (2) Using the output ofocclusion detection, this project will further propose an occlusion handling scheme for light field datawith reflections from transparent glasses. (3) This project will propose a disparity estimation scheme,especially for textureless regions in the light field data with reflections from transparent glasses. Theresearch results of this project are expected to increase accuracy of reconstructed depth for light fielddata with reflections, and thus improve the output image quality of image rendering.