Abstract
Keypoint matching between images is an important technique for computer vision applications such as image retrieval. Although binary feature descriptors such as BRIEF enable fast measurement of distance, exhaustive search is still time-consuming. Hashing methods such as Locality Sensitive Hashing (LSH), while being effective to accelerate searching, result in large memory consumption and thus are difficult to apply in some cases, such as where keypoint matching is implemented on a Field Programmable Gate Array (FPGA). On the other hand, Norm Ordered Matching (NOM), in which feature descriptors are ordered by their norm and searched in the scope of which the norm is close, consumes almost no extra memory. However, it has a problem in that the norm of a binary feature descriptor tends to concentrate around half of the length of the feature descriptor, which leads to large computation time. In this paper, we propose a method, called ENOM, which equalizes the distribution of norms with no impact on distances between feature descriptors by flipping bits on some common locations of feature descriptors to accelerate searching. Experimental evaluation shows that ENOM is faster than NOM, and it is equivalent in precision and speed to LSH as well as being superior in terms of memory consumption.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.