%0 Journal Article
%T Image Interpolation and Error Concealment Scheme Based on Support Vector Machine<br>基于支持向量机的图象插值及错误隐匿策略
%A WANG Jue
%A JI Liang
%A <br>王珏
%A 季梁
%J 中国图象图形学报
%D 2002
%I 
%X How to prevent quality degradation due to channel errors for images and video transmitting over lossy channels is a fundamental problem in multimedia signal processing. Support Vector Machine(SVM) is a novel powerful learning method and is now a new hotspot in the field of machine learning, and has been successfully used in many pattern recognition problems. To get more satisfying error concealment results, a novel error concealment scheme based on SVM is proposed in this paper. At first, a SVM based image interpolation algorithm is successfully established. SVM learning machines are carefully trained by a large amount of training data exacted from standard images to learn the relationship of neighboring pixels in spatial domain, and theses well trained machines are used as specific nonlinear interpolation operators. Comparative results show that this kind of interpolation operator outperforms not only some traditional used interpolation operators such as linear and median operators, but also some operators carefully trained by artificial neural networks. The error concealment problem is placed into a spatial image interpolation framework and the proposed interpolation method is thoroughly used to estimate the missing image blocks according to their neighboring pixels. Experimental results show that compared with some error concealment schemes both in spatial and frequent domain in the literature, especially those based on artificial neural networks, the proposed one has remarkable advantages in error concealment performance and generalization property.
%K Error concealment
%K Support vector machine
%K Nonlinear interpolation<br>图象插值
%K 错误隐匿
%K 支持向量机
%K 非线性插值
%K 机器学习
%K 网络环境
%K 视频传输
%U http://www.alljournals.cn/get_abstract_url.aspx?pcid=5B3AB970F71A803DEACDC0559115BFCF0A068CD97DD29835&cid=8240383F08CE46C8B05036380D75B607&jid=D06194629680C940ACE75262F54B9D85&aid=C5E18A51F9510EE4&yid=C3ACC247184A22C1&vid=DF92D298D3FF1E6E&iid=B31275AF3241DB2D&sid=C7A2B92569DF5458&eid=2497388423811B81&journal_id=1006-8961&journal_name=中国图象图形学报&referenced_num=0&reference_num=9