| Summary: | 碩士 === 義守大學 === 資訊工程學系 === 104 === In this thesis, we will develop an effective video logo detection method which use the characteristics of logo in various video programs.
For video logo, we can roughly divide into three different kinds, opaque, translucent and motional. Generally, we can use the approaches of pixels difference or average grey level to detect the opaque logo. However, for translucent or motional logo, they cannot detect effectively.
In our work, we propose a gradient-based method to detect video logo. The new method can apply to detect opaque or translucent logo effectively. For motional logo, the method can apply directly due to the insignificant motion. On the other hand, in this thesis we also extensively study the properties of the logo that appear in various video. Therefore, the new method is robust for most video and achieves satisfactory results.
|
|---|
