【作者】 于佳； 王燕；

【Author】 YU Jia,WANG Yan (School of Computer Science and Technology,Harbin Engineering University,Harbin 150001,China)

【机构】 哈尔滨工程大学计算机科学与技术学院；

【摘要】 提出了一种基于有导师学习的SOM神经网络的视频中人工文本区域检测方法。SOM神经网络具有自组织、自适应的学习能力,在基于灰度统计特征的基础上,采用SOM分类器,针对视频中的人工文本区域进行检测。现有的方法大多数都受到了文字形状以及出现的位置等一些因素的制约,提出的方法在一定程度上解决了这个问题。实验结果表明,该方法能够有效地检测和定位视频中的人工文本区域。更多还原

【Abstract】 A detection approach of artificial text area in video based on supervised learning.SOM was proposed. SOM has the ability of self-organizing and self-adapting,used SOM as a classifier based on the statistic feature of gray level to detect the artificial text area in video.And the current methods were all effected by the shape and the location of text,the method which was proposed resolved this problem efficiently.The result of the experiment showed that this kind of method was a very effective way to detect and locate the artificial text area in video.更多还原