Accuracy and performance analysis of time coherent 3D animation reconstruction from RGB-D video

Abstract

We present an accuracy and performance analysis of Time Coherent 3D Animation Reconstruction methods from RGB-D video. We analyze the existing methods that can reconstruct a time coherent 3D animation using RGB-D video. We also present a modified algorithm using only the RGB data that extends the analysis of existing methods. We show that using all the methods it is possible to reconstruct a time-coherent 3D animation using either only the color data, color and depth data, or only the depth data. We compare all the methods using a number of error measures and analyze the strength and weaknesses of each method in terms of their accuracy and runtime performance. Our analysis demonstrates that given RGB-D video data, it is possible to select the best algorithm for time coherent 3D animation reconstruction under a number of constraints in terms of the required accuracy and runtime performance.