V.V. Voronin, V.I. Marchuk, V.А. Frantc, N.V. Gapon
Currently, the decision was widespread applications processing and video analytics systems in computer vision. Most of them are done manually by the user, in connection with this is especially true solution automatically restore the video sequence, and delete objects. In the video there are static images that obscure the video, covering part of the useful information from the viewer. These images include various logos channel, date, time, or subtitles superimposed on the video signal with the further coding. In addition, the disturbing factor is the distortion power at work video codec, which is explained by the appearance of the unreliability of the data channel from the encoder to the decoder.In some cases, there may be unwanted objects on the frame sequence. An object is understood to be a connected region of pixels per frame, considered as a whole. The object can be a moving car or a person, the defect caused by a scratch on the film or the entire background scene.
This paper describes a new approach to the restoration of the video, which is based on getting the model scene and tracking of moving objects. The proposed approach allows you to remove objects or restore the missing or damaged area of the video sequence using spatial and temporal information of adjacent frames. The algorithm iteratively performs the following operations: receiving a frame, update the model scene, updating the position of moving objects, the replacement of the frame with objects marked for deletion. As a possible model of the background considered moving average code book and several others. The main advantage of this approach is the ability to operate in streaming mode.