A 3D reconstruction from real-time stereoscopic images using GPU

Jose-Ernesto Gomez-Balderas (1) , Dominique Houzet (1)

1 GIPSA-AGPIG - GIPSA - Architecture, Géométrie, Perception, Images, Gestes

In this article we propose a new technique to obtain a three-dimensional (3D) reconstruction from stereoscopic images taken by a stereoscopic system in real-time. To parallelize the 3D reconstruction we propose a method that uses a Graphics Processors Unit (GPU) and a disparity map from block matching algorithm (BM). The results obtained permit us to accelerate the images processing time, measured in frames per second (FPS) with respect to the same method using a Central Processing Unit (CPU). The advantage of speed using GPU advocates our system for practical applications such as aerial reconnaissance, cartography, robotic navigation and obstacle detection.

Mots clés: real-time; Graphics Processing Units (GPU); stereoscopic images; block matching; 3D reconstruction

Domaines: Vision par ordinateur et reconnaissance de formes [cs.CV]

Jose-Ernesto Gomez-Balderas, Dominique Houzet. A 3D reconstruction from real-time stereoscopic images using GPU. DASIP 2013 - Conference on Design and Architectures for Signal and Image Processing, Oct 2013, Cagliari, Italy. pp.253-258. ⟨hal-00878683⟩