Title :
A Pyramid Nearest Neighbor Search Kernel for object categorization
Author :
Hong Cheng ; Rongchao Yu ; Zicheng Liu ; Yiguang Liu
Author_Institution :
Univ. of Electron. Sci. & Technol. of China, Chengdu, China
Abstract :
Nearest-Neighbor based Image Classification (N-NIC) has drawn considerable attention in the past several years because it does not require classifier training. Similar to an orderless Bag-of-Feature image representation, the traditional NNIC ignores global geometric correspondence. In this paper, we present a technique to exploit the global geometric correspondence in a nearest neighbor classifier framework. We divide an image into increasingly fine sub-regions like the Spatial Pyramid Matching (SPM) approach, and introduce a Pyramid Nearest Neighbor Search kernel by measuring the search similarity between a local descriptor and a feature set in each pyramid window. Instead of using a fixed weighting as in SPM, the weights of the pyramid windows are learned in a class-dependent manner. By doing so, we learn a class-specific geometric correspondence. Finally, an optimal nearest neighbor classifier framework is developed to incorporate the kernel functions over different pyramid windows. We evaluate our proposed approach on a number of public datasets, and show the results significantly outperform existing techniques.
Keywords :
computational geometry; image classification; image matching; set theory; NNIC; SPM; class-specific geometric correspondence; feature set; global geometric correspondence; local descriptor; nearest neighbor classifier framework; nearest-neighbor based image classification; object categorization; orderless bag-of-feature image representation; pyramid nearest neighbor search kernel; pyramid windows; search similarity measurement; spatial pyramid matching; Accuracy; Educational institutions; Kernel; Nearest neighbor searches; Testing; Training; Vectors;
Conference_Titel :
Pattern Recognition (ICPR), 2012 21st International Conference on
Conference_Location :
Tsukuba
Print_ISBN :
978-1-4673-2216-4