Similar Scene Classification Research Based on Dense Matching
Scene classification is one of the important topics of computer vision, and the classification of similar scenes is even more challenging. This paper proposes a new method for image representation suitable for such a task. First, a displacement vector map of an input scene image can be obtained by utilizing SIFT-Flow. Then, the map is segmented into spatial blocks, so that information about the matching result can be used for creating a representation for the image. Finally, scene images can be classified by Supported Vector Machine (SVM). The proposed method outperforms state-of-the-art approaches for classifying similar scenes.
Image representation SIFT-Flow displacement vector map SVM Scene classification
Han Chao Hou Jianjun Xu Lingqing Bai Shuang
School of Electronic and Information Engineering, Beijing Jiaotong University, Beijing, China
国际会议
重庆
英文
235-239
2015-12-18（万方平台首次上网日期，不代表论文的发表时间）