Wide Baseline Matching [chapter]

Tinne Tuytelaars
2014 Computer Vision  
Wide baseline matching is the state of the art for object recognition and image registration problems in computer vision. Robust feature descriptors can give vast improvements in the quality and speed of subsequent steps, but intensive computation is still required. With the release of general purpose parallel computing interfaces, opportunities for increases in performance arise. In this paper we present an implementation of Speeded-Up Robust Feature (SURF) extractor, based on the OpenCL
more » ... on the OpenCL system of GPU programming developed by NVIDIA. For an 800x640 pixel image, the GPU-based method executes nearly 5 times faster than a comparable CPU-based method, with no significant loss of accuracy.
doi:10.1007/978-0-387-31439-6_191 fatcat:qarlzawgt5hb5of4ccmxqhikl4