Researches On Methods For Video Logo Detection And Inpainting

Logo is an identifying symbol in digital video. The presence of logo impacts the visual effect and spread of video. Therefore, how to remove the logo quickly and effectively has become a very important problem. The development of inpainting technology for image and video offered a possible method to solve this problem. In this paper, several classical image and video inpainting algorithms are reviewed. By the help of analyze the advantages and disadvantages between these algorithms, we studied the method of removing the video logo by image and video inpainting technology.Firstly, according to texture and structure characters of immobile logo, a novel coarse-to-fine logo detection algorithm based on SVM is presented. At first, the changes between video frames are calculated and the candidate logo is segmented. Then the logo region can be accurately detected and located by SVM. Experiments show that this method can detect and locate transparent or semi-transparent logo effectively.Secondly, considering that current image inpainting algorithms are ineffective for large-scale missing region, we present a novel image inpainting algorithm based on wavelet transform. By making full use of the wavelet coefficients, the inpainting process is greatly sped up and the inpainted results are more visually plausible. The experimental results prove that our method is more robust, faster and more practicable.Then, to remove immobile logo and keep high spatial-temporal quality of the logo-removed video, a novel video inpainting algorithm based on Poisson equation is presented. At first, the edge structure of video is restored through the global optimization procedure. Then a Poisson equation is set based on the restored structural information, and video is inpainted by solving this Poisson equation.Finally, the conclusions summarized the work of this paper and pointed out the work needing to deal with in the further research.