Background Inpainting Considering Scene Characteristics of Video

In recent years, video inpainting has been widely used in many applications such as error correction in video, multimedia editing, secrecy protection, and subtitle removal. Especially, video inpainting in computer-aided film production has been extensively studied for various purposes such as scene synthesis. The problem with video inpainting is to finish the treatment in a practical time while getting satisfactory inpainting results. There are some short run-time methods, but they also have serious defects such as not guaranteeing temporal consistency of video. In reality, studying practical high-speed video inpainting methods to ensure temporal consistency in video is a pressing problem. On the other hand, many applications require background inpainting in video. In this paper, we propose a method of inpainting of backgrounds that can be executed quickly by extending the priority-based image restoration method to video. By applying appropriate inpainting methods according to the scene characteristics of the video, the quality of the inpainting is guaranteed similar to the global search method.

Autorentext

Ha Jong Won received the M.Sc. degree in intelligent information processing in 2006, Ph.D. degree in visual information processing in 2016 from Kim Il Sung University, Pyongyang, DPR of Korea.He is currently an Associate Professor with the College of information Science, Kim Il Sung University, Pyongyang, DPR of Korea.