一种基于二值图角点匹配的台标识别方法  

TV Logos Recognition Method Based on Binary Image Corner Matching

作　　者：金阳[1] 程江华[1] 任通[1] 库锡树[1] 

机构地区：[1]国防科技大学电子科学与工程学院,湖南长沙410073

出　　处：《电视技术》2012年第17期15-19,29,共6页Video Engineering

摘　　要：电视台标在视频传输、保存和格式转换过程中易发生缩放、旋转等形变,且向半透明化方向发展,这些因素给台标的识别带来了困难。针对形变及半透明化问题,提出一种基于二值化图像角点匹配的三步骤台标识别方法。首先利用二值化台标图像的差分累加和求平均,得到分割后的实时台标图像;然后利用Susan算法不涉及梯度运算的特点,在二值化台标图像上提取角点;最后采用极坐标域角点对匹配的方法,实现台标匹配。使用了部分地方电视台台标进行实验,结果证明了该方法的有效性。Because of transmission, preservation, format conservation and so on, logos always have problems of transformation and circumrotation. It is more difficult to recognize a logo while the logo tend to semi-transparent. A TV logo recognition method is proposed to solve these problems which based on binary image corner matching by three-step process. Firstly, the real-time TV station logo image is obtained by using the average of several accumulated difference sums of the segmented binary TV station logo images. Secondly, corners is extracted based on segmented binary TV station logo images by using the Susan arithmetic, which has a feature of no gradient operation; Finally, the logo is matched after working out the matching corners in transformed polar coordinates. By using several local TV logos, the experiments＇ result indicate that this method is effective.

关 键 词：台标识别 图像分割 角点提取 角点匹配 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]