Accelerating image local invariant features matching using cloud computing platform

Local invariant feature is a powerful tool for finding correspondences between images. However, current feature descriptors typically suffer the lack of global context and fail to resolve ambiguities that can occur when an image has multiple similar regions. Meanwhile, local features matching suffers expensive time cost, and cannot meet the needs of real time applications. This paper encompasses a global descriptor and architecture of cloud computing platform for increasing both the accuracy and efficiency of the well known SIFT method. Firstly, our global descriptor makes full use of the global characteristics of initial matched feature points obtained by SIFT for filtering out mismatches. Next, we will speed up SIFT using our cloud computing platform, iVCE. Experimental results have demonstrated that the global descriptor proposed can effectively improve the accuracy of the matched feature points without reducing the performance. The matching speed can also be enhanced dramatically by executing SIFT on our cloud computing platform.