Fast and Texture-Structure Preservable Inpainting and Its Application for Creating Virtualized-Reality Indoor Models

This paper discusses the new inpainting algorithm for handling the images in the virtualized-reality indoor models. The untextured regions in the indoor model are textured by this proposed inpainting method which guarantees the texture and the structure propagation. We call this as a hybrid method because it combines the effectiveness of the two popular inpainting algorithms namely, the Exemplar-Based method and the Edge-Based image restoration method and rectifies the drawbacks in these two methods. The tedious patch search in the Exemplar-Based method is simplified by the usage of the hash table which holds the classified textures. The Edge-Based method can handle the structures only in the form of lines and circles whereas the proposed method can handle any type of curves by using the combination of superpixels and spline curves.