Flexible Image Similarity Computation Using Hyper-Spatial Matching
Jianxin Wu
Assistant Professor
School of Computer Engineering, Nanyang Technological Univeristy
Abstract: Spatial pyramid matching (SPM) has been widely used to compute the similarity of two images in computer vision. In order to achieve the best match, SPM implicitly assumes that similar objects appear in the corresponding locations in two images of the same category. However, this is not always the case. In this paper, we propose hyper-spatial matching (HSM), a more flexible image similarity computing method, to overcome the mis-matching problem in SPM. Besides the match between corresponding regions, HSM considers the relationship of all spatial pairs in two images, which includes more meaningful match than SPM. We propose two learning strategies to learn SVM models with the proposed HSM kernel in image classification, which are hundreds of times faster than a general purpose SVM solver applied to the HSM kernel (in both training and testing). We compare HSM and SPM on four large scale benchmarks including the SUN dataset, and show that HSM is better than SPM in describing the image similarity.
Bio: 吴建鑫，分别于1999年7月和2002年7月在南京大学计算机科学与技术系获学士和硕士学位，并于2009年8月在美国佐治亚理工学院获计算机科学专业的博士学位。2009年8月起在新加坡南洋理工大学计算机工程学院任助理教授，2012年依托南京大学入选中组部第三批青年千人计划。目前主要从事计算机视觉和机器学习的研究工作，研究的主要问题是场景理解，包括场景的识别以及场景中人、物体、行为的检测和识别等。具体的研究兴趣包括图像特征提取（如何从图像和视频中提取有效的特征以进行识别和检测）和大规模数据的学习（如何使用百万张以上图像和其他数据高效、精确地学习模型）。