Image Inpainting based on Posterization
| Autor: | Hsiao-Han Lo, 羅筱涵 |
|---|---|
| Rok vydání: | 2007 |
| Druh dokumentu: | 學位論文 ; thesis |
| Popis: | 95 Image inpainting is an interesting and practical image processing technology. We can remove the object in the image at our pleasure, and fill in the removed pixels (i.e. damaged area) by image inpainting. For the purpose of patching up automatically, filling in large-sized damaged areas in the image, and lowering down the operation complexity, we choose the exemplar-based image inpainting as our main framework. In order to keep the structure consistency between the damaged areas and the non-damaged areas, we redefine the two factors, Confidence and Structurality, which determine the filling order of target (damaged) patches. With these two newly defined terms, we propose a dynamic rule to choose the best target patch and the suitable patch size in each filling step. In addition, we also reduce the searching region of source patches by posterization, thus leading to a faster speed of impainting. |
| Databáze: | Networked Digital Library of Theses & Dissertations |
| Externí odkaz: |
|