THTM: A template matching algorithm based on HOG descriptor and two-stage matching

We propose a novel method for template matching named THTM – a template matching algorithm based on HOG (histogram of gradient) and two-stage matching. We rely on the fast construction of HOG and the two-stage matching that jointly lead to a high accuracy approach for matching. TMTM give enough attention on HOG and creatively propose a twice-stage matching while traditional method only matches once. Our contribution is to apply HOG to template matching successfully and present two-stage matching, which is prominent to improve the matching accuracy based on HOG descriptor. We analyze key features of THTM and perform compared to other commonly used alternatives on a challenging real-world datasets. Experiments show that our method outperforms the comparison method.We propose a novel method for template matching named THTM – a template matching algorithm based on HOG (histogram of gradient) and two-stage matching. We rely on the fast construction of HOG and the two-stage matching that jointly lead to a high accuracy approach for matching. TMTM give enough attention on HOG and creatively propose a twice-stage matching while traditional method only matches once. Our contribution is to apply HOG to template matching successfully and present two-stage matching, which is prominent to improve the matching accuracy based on HOG descriptor. We analyze key features of THTM and perform compared to other commonly used alternatives on a challenging real-world datasets. Experiments show that our method outperforms the comparison method.

[1]  Patrick Pérez,et al.  Color-Based Probabilistic Tracking , 2002, ECCV.

[2]  Huchuan Lu,et al.  Visual tracking via adaptive structural local sparse appearance model , 2012, 2012 IEEE Conference on Computer Vision and Pattern Recognition.

[3]  Shai Avidan,et al.  Locally Orderless Tracking , 2012, 2012 IEEE Conference on Computer Vision and Pattern Recognition.

[4]  Bill Triggs,et al.  Histograms of oriented gradients for human detection , 2005, 2005 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR'05).

[5]  Tomoko Ozeki,et al.  Pedestrian Detection by Using FAST-HOG Features , 2015, HAI.

[6]  Trevor Darrell,et al.  Rich Feature Hierarchies for Accurate Object Detection and Semantic Segmentation , 2013, 2014 IEEE Conference on Computer Vision and Pattern Recognition.

[7]  Chao Zhang,et al.  Two-Side Agreement Learning for Non-Parametric Template Matching , 2017, IEICE Trans. Inf. Syst..

[8]  Lihi Zelnik-Manor,et al.  Template Matching with Deformable Diversity Similarity , 2016, 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

[9]  Alexander Sibiryakov,et al.  Fast and high-performance template matching method , 2011, CVPR 2011.

[10]  Leonidas J. Guibas,et al.  The Earth Mover's Distance as a Metric for Image Retrieval , 2000, International Journal of Computer Vision.

[11]  Clark F. Olson.  Maximum-Likelihood Image Matching , 2002, IEEE Trans. Pattern Anal. Mach. Intell..

[12]  Federico Tombari,et al.  Performance Evaluation of Full Search Equivalent Pattern Matching Algorithms , 2012, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[13]  William T. Freeman,et al.  Best-Buddies Similarity for robust template matching , 2015, CVPR.

[14]  Michael Werman,et al.  Robust Real-Time Pattern Matching Using Bayesian Sequential Hypothesis Testing , 2008, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[15]  Haibin Ling,et al.  Real time robust L1 tracker using accelerated proximal gradient approach , 2012, 2012 IEEE Conference on Computer Vision and Pattern Recognition.

[16]  Kang-Hyun Jo,et al.  Hybrid cascade boosting machine using variant scale blocks based HOG features for pedestrian detection , 2014, Neurocomputing.

[17]  Denis Simakov,et al.  Summarizing visual data using bidirectional similarity , 2008, 2008 IEEE Conference on Computer Vision and Pattern Recognition.