Broadcast Video Logo Detection and Removing
| Autor: | Jin-Long Shen, 沈金龍 |
|---|---|
| Rok vydání: | 2007 |
| Druh dokumentu: | 學位論文 ; thesis |
| Popis: | 95 Broadcast video logos can be divided into three types which are opaque, translucent and animation logos. The opaque logos can be segmented by using the pixel-based difference to compute the gray-level average, but the method does not achieve satisfactory performance for the translucent logos. In the thesis, we use the gradient-based average computation to detect the logos. It is effective to detect the opaque logos and the translucent logos. The logo regions are removed, and then be recovered by the image-inpainting technique. The image-inpainting technique can be used to repair damaged regions and remove objects or textures in images. We use two algorithms, texture synthesis and inpainting, to deal with the the inpainting problem. The exemplar-based image inpainting, that uses the simulation way, combines the two algorithms to remove large objects and fill in small holes. In the thesis, we propose an algorithm that can detect broadcast video logos automatically, and use the image inpainting technique to remove the broadcast video logos. Experimental results indicate that the proposed method efficiently to detect and remove the broadcast video logos. |
| Databáze: | Networked Digital Library of Theses & Dissertations |
| Externí odkaz: |
|