Abstract

Large-scale image-based modeling is a challenging and important research topic, which has wide applications in various areas. Existed serial methods consume a significant amount of time for large problems. With the emergence of multi-core computers, it is possible to speed up the method using central processing unit (CPU) parallelism. In this paper, we present the design and implementation of multi-core large-scale image-based modeling that exploits hardware parallelism to efficiently solve the large-scale 3-D scene reconstruction problem. We explore the use of multi-core CPU to achieve high speedups for the whole problem. The speedup ratio can be close to the number of cores. Experimental results show that our method achieves the acceleration of the whole algorithm and also ensures the accuracy of the 3-D reconstruction.

Key words: Large-scale image-based modeling, multi-core, central processing unit (CPU) parallelism, 3-D reconstruction, speedup ratio.