Seminar_gy.pptVIP

Image inpaintingImage Completion with Structure PropagationJian Sun LuYuan Jiaya Jia ACM Transactions on Graphics, 2005Graphcut Textures: Image and Video Synthesis Using Graph CutsVivek Kwatra Arno Schodl Irfan EssaACM Transaction on Graphics, SIGGRAPH 2003 reporter: Gui Yan 2008/11/7 Outline Image inpainting 1.What is the Image inpainting 2.Related work 3.Observations Structure Propagation 1.The Concept of Structure Propagation 2. Energy Minimization 3. Dynamic Programming (DP) Texture synthesis 1. The concept of texture 2.Graph Cuts 3. Patch Placement Matching What is the Image inpainting Image inpainting is filling in missing pixels in a large unknown region of an image in a visually plausible way. I : input image Ω: unknown or missing region Related work 1.Image inpainting, [introduced by Bertalmio et al. 2001] works at the pixel level, and works well for small gaps, thin structure, and so on; 2.Example-based approaches [Igehy and Pereira 1997; Efros and Freeman 2001 et al] use texture synthesis techniques with automatic guidance; 3.Interactive guidance [pérez et al 2004], using depth information, and “point of interest” to improve inpainting results. Problems These methods have difficulties completing images where complex salient structures exist in the missing region. (Such salient structures may include curves, T-junctions, and X-junctions.) Previous patch-by-patch completion algorithms operate in a greedy manner that may also cause discontinuities in salient structures. Observations The most salient missing structures can be approximated by a few well-defined curves. There exists a synthesis ordering for image inpainting: the regions with salient structures should be completed before filling in other regions. The approach proceeds in three step