Files in this item



application/pdfECE499-Sp2012-ma.pdf (779kB)Restricted to U of Illinois
(no description provided)PDF


Title:Depth Video Enhancement on FPGAS
Author(s):Ma, Sai
Contributor(s):Chen, Deming
Subject(s):computer vision
three-dimensional computer vision
depth video
field-programmable gate arrays
Abstract:Providing high-quality depth data has been one of the most important issues in the field of 3D computer vision. In contrast to a number of methods on fusion of stereo and active depth sensors, a novel approach to generate depth video with one color video and its corresponding low-resolution depth video has been proposed [1]. Though the quality of the enhanced depth video has been improved, the real-time requirement still remains as a crucial challenge because of intensive computations. With the time profiling run by AMD codeanalyst, the floating-point calculations in the up-sampling helper functions take up to 75% of the total running time. So, to meet the demand of highly complex calculations, FPGAs have become desirable platforms due to their reconfigurable logic blocks, distributed memory and DSP modules. Therefore, in this thesis, we implement a system in reconfigurable hardware using efficient parallel architecture and impulse CoDeveloper, a C-to-FPGA tool. The designed enhancement method exploits hardware parallelism to achieve high system speed. For the hardware implementation, we use Xilinx series FPGA device. The proposed FPGA-based method is compared with the software equivalent and the observed speedup is about seven-fold.
Issue Date:2012-05
Publication Status:unpublished
Peer Reviewed:not peer reviewed
Date Available in IDEALS:2014-01-09

This item appears in the following Collection(s)

Item Statistics