| Summary: | 碩士 === 國立中正大學 === 資訊工程所 === 97 === Image inpainting is to remove unwants objects or reconstruct damaged parts of an image automatically, which is an ill-posed problem in image processing.
In this study, an image inpainting approach using edge-guided belief propagation (BP) and graph cuts is proposed. The proposed image inpainting approach contains five stages, i.e., determine MRF nodes, enlarge the label set by various geometric transformations, extract the edge map of the image, image inpainting using priority-BP optimization, and compose each overlapped region between two neighboring MRF nodes. Based on experimental results obtained in this study, as compared with three comparison methods, the proposed approach provides the better image inpainting results.
|
|---|
