Fast Hash-Based Inpainting for Virtualized-Reality Indoor Modeling

摘要

This paper discusses the hash table usage for the Exemplar-Based inpainting method to speed up the inpainting process. The time consuming stage in the Exemplar-Based inpainting is the texture patch searching process where the occluded texture patch has to be compared with all other texture patches available in the rest of the image. The proposed hash function reduces the patch search into single/minimum search. Each texture patch is given an unique location in the hash table so that they are easily picked up by their addresses during every iteration. Gray level co-occurrence matrix (GLCM) is used for designing the hash function. The proposed hash works are added to the Exemplar-Based inpainting and this new hash-based inpainting method is tested in the planes of the virtualized-reality indoor model. There are often un-textured regions or the distorted textures in the 3D planes of the virtualized-reality model which needs inpainting. Applying the proposed fast hash-based inpainting, works for the efficient post processing in the virtualized-reality indoor modeling. This removes the necessity for the users to capture the left out scenes for the hidden regions. The proposed fast hash based inpainting reduces the computation time and provides good quality textures. The novelty of this paper lies in the application of the GLCM matrix in designing the hash function and the occlusion handling measures.