Character Superimposition Inpainting in Surveillance Video

Video surveillance systems play an important role in the crime scene investigation, and the digital surveillance system always requires the superimposed video data being subjected to a data compression processing. The purpose of this paper is to study the use of inpainting techniques to remove the characters and inpaint the target region. We give the efficient framework including getting Character Superimposition mask, superimposition movement and inpainting the blanks. The character region is located with the manual ROI selection and varying text extractor, such as the time. The superimposed characters usually have distinguished colors from the original background, so the edges are easily detected. We use the canny operator the get the edge image. The missing information which is effect the structure of the original image is reconstructed using a structure propagating algorithm. The experiment was done with C/C++ in the vs2010 KDE. The framework of this paper showed is powerful to recreate the character superimposition region and helpful to the crime scene investigation.