Authors: Yuxiang Xie XiaoPing Zhang Xidao Luan Li Liu Xin Zhang
Publish Date: 2013/05/28
Volume: 74, Issue: 1, Pages: 105-122
Abstract
Automatic image scene detection is a crucial step for various tasks in computer vision Current scene detection methods are often computationally expensive for use in realtime image classification In this paper a novel and efficient scene detection method based on local invariant features is presented First the SIFT feature detector and descriptor has been utilized to extract local image features since the SIFT descriptor has been proved to be an excellent local method that yields high quality features However the SIFT descriptor has been shown to produce high dimensional and redundant local features which can create processing difficulty and computational burden in the successive classification stage Therefore two new feature selection strategies are proposed to reduce the number of SIFT keypoints and hence reduce the computational complexity In both strategies each image is represented by a single feature vector which assures the efficiency Finally a multiclassifier based on a support vector machine is applied to perform the scene detection task Experimental results show that the proposed method can achieve accurate satisfactory classification results with significantly reduced computational complexityThis work has been supported by the National Natural Science Foundation of China under contract No 61201337 and by Changsha Municipal Science and Technology Project under contract No K120504511 The authors are grateful to the anonymous reviewers for valuable comments
Keywords: