CPU and GPU parallel processing for mobile Augmented Reality

This paper introduces a parallel processing scheme using CPU and GPU for the mobile Augmented Reality applications. Most of AR applications have to perform intensive image processing algorithms to detect specified objects on which virtual information is overlaid. The object detection generally consists of a feature extraction module and a feature description module. The proposed scheme distributes the feature extraction module and the feature description module to CPU and GPU respectively, and performs the modules in parallel. In experimental results, the proposed scheme outperforms the CPU only scheme and the sequential CPU-GPU execution scheme.