Abstract: High-definition 3D video for real-time applications is one of the features explored by the next generation of telecommunication systems. From an implementation point for view, the most restrictive limitations of previously published methods are processing speed and memory bandwidth. In this paper we present several algorithmic improvements for Visual Hull (VH) reconstruction using a voxel-based approach that reduces the resource consumption. We also present FPGA and GPU based implementations. These improvements have been shown to outperform processing speeds of the latest publications, allowing real-time and high-definition implementations. The proposed approach allows a 256 × 256 × 128 reconstruction volume to be obtained in only 33.55 ms in a FPGA.

Autores/as

Descargar referencia