旋转与尺度不变模板匹配的快速筛选算法
人机交互
2017-07-19 v1
摘要
本文提出了一种通用的预处理器，用于加速传统模板匹配技术。所提出的算法并非通过穷举搜索在参考图像中定位与查询模板最佳匹配的图块，而是以最小的计算量排除不可能存在匹配的区域。虽然仅使用均值、方差和梯度等简单的图块特征，但该快速预筛选具有高度的判别力。其计算效率得益于使用了一种新颖的八角星形模板，并利用容斥原理来提取和比较图块特征。此外，它能有效处理参考图像的任意旋转和缩放。大量实验表明，所提出的算法极大地减少了搜索空间，同时从未遗漏最佳匹配。
引用
@article{arxiv.1707.05645,
title = {Prototype Tasks: Improving Crowdsourcing Results through Rapid, Iterative Task Design},
author = {Snehalkumar "Neil" S. Gaikwad and Nalin Chhibber and Vibhor Sehgal and Alipta Ballav and Catherine Mullings and Ahmed Nasser and Angela Richmond-Fuller and Aaron Gilbee and Dilrukshi Gamage and Mark Whiting and Sharon Zhou and Sekandar Matin and Senadhipathige Niranga and Shirish Goyal and Dinesh Majeti and Preethi Srinivas and Adam Ginzberg and Kamila Mananova and Karolina Ziulkoski and Jeff Regino and Tejas Sarma and Akshansh Sinha and Abhratanu Paul and Christopher Diemert and Mahesh Murag and William Dai and Manoj Pandey and Rajan Vaish and Michael Bernstein},
journal= {arXiv preprint arXiv:1707.05645},
year = {2017}
}
备注
2 pages (with 2 pages references, 2 pages Appx), HCOMP 2017, Association for the Advancement of Artificial Intelligence (www.aaai.org)