The Spectrum Broadening in the Plenoptic Function

Abstract

The spectral analysis of the plenoptic function is an intuitive way to acquire the sampling rate in the image-based rendering (IBR) data. The bandwidth of the spectrum directly determines the minimum sampling rate and the reconstruction filter. In this paper, we investigate the spectrum broadening of some complicated scenes including the non-Lambertian reflection scene, the occlusive scene and a slanted plane scene. By modeling the EPI lines in the maximum and minimum depths, the broad range is analyzed. The spectrum of plenoptic function is constructed by two parts. Firstly, the maximum and minimum depths build the main part of the plenoptic function spectrum. Secondly, factors such as the material property, the depth difference, the scene slope and the texture information form the side lope of the spectrum which broadens the spectrum. Lastly we derive the new Nyquist rates for the plenoptic sampling of such complex scenes and new reconstruction filters. Simulation results demonstrate the effectiveness of our approach.