Restoration of Frame Borders under Stabilization based on Background Model Building and Salient Objects Estimation
Abstract
Introduction: Digital video stabilization is often used in video processing systems and video surveillance in order to improve a quality of final video sequences. One of the main challenges deals with a loss of border regions of a frame during its scaling. The investigations in a framework of the alternative approaches for frame border restoration did not lead to reasonable algorithmic solution providing the required quality of frame. Purpose: The goal is to develop an effective algorithm providing a frame border restoration under video stabilization based on the background model, estimation of the salient objects in a frame, and multi-band blending the border frame and reconstructed regions. Results: An algorithm has been developed for frame border restoration after stabilization stage. It is based on splitting the scene into a background model and salient objects, forming an intermediate buffer frame and stitching the frame borders for video sequences including a non-uniform movement of the camera and objects, as well as a multilevel scene. Practical relevance: The results were tested using a common dataset Sports Videos for Wild with a large number of video sequences obtained by a moving camera, and also video sequences L1 Robust Optimal Camera Path commonly used as video stabilization test material. The implementation of the developed algorithm allows the decrease in the frame scaling and frame losses down to 3-5% even in the cases of complex movement and complex structure of a scene.Published
2017-10-20
How to Cite
Buryachenko, V., Favorskaya, M., Zotin, A., & Pahirka, A. (2017). Restoration of Frame Borders under Stabilization based on Background Model Building and Salient Objects Estimation. Information and Control Systems, (5), 42-51. https://doi.org/10.15217/issn1684-8853.2017.5.42
Issue
Section
Information processing and control