Abstract: A framework for playing the missing parts of a movie recorded with a moving or stationary camera is presented in this document. The area to be retouched is general: it can be still or animated background or foreground object that can block and be blocked by another object. The algorithm is a simple pretreatment step and two step video editing. In the pre-processing stage, each image is the segment in the background. Segmentation is used to build three image mosaics that help produce results in constant time and also improve the performance of the algorithm re-produce the search space. In the video inpainting step first, the reconstruction of moving objects in the foreground that "occlusion" in the region to be retouched. For this, the gap filled as much as possible by copying information moves in other parameters, using a system based on priorities. In the second step, Inpaint the remaining hole with the background. To do this, the first line and copy images directly as possible. The remaining pixels filled by extending spatial texture synthesis techniques in the spatial-temporal. The proposed framework has several advantages over the state of the art algorithms that deal with the same data types and constraints. It gives some camera movements are simple to use, fast, does not require statistical models in the background or foreground, works well in the presence of rich and cluttered backgrounds and the results show that there is no 'visible artifacts or motion blur. Several examples are presented to support these conclusions.
Download Link :
http://pptpapers.googlegroups.com/web/Video+Impainting+using+Constrained+Camera+Motion.doc?gda=7DHbM2cAAACw9PVFGg6n8BB4EzqJqwTrK1M5_Ja_ggp-5w2x8-T-sdKP6PF7egjgljDfehL6MRZBzWYI3NULP18mS8_7wPq2QTHyH3rDTJKBA9Pok9n8j91c7U-vCZ5O0GVCYZoXMkp5Xh26_qkMwaGGFl2NCU0D