Study On The Techniques Of Flicker Removal Based On Nonlinear Model

As the rapid development of communication technology, computer network technology, multimedia processing technology, the application prospect of video surveillance system is more and more widespread. Flicker is inevitably caused by varied light environment and the auto exposure function of video cameras in video capture process. Flicker can easily make viewers visual fatigue, and affect the quality of monitoring video. Moreover, it increases the video redundancy in the transmission, which is not conducive to data stream real-time transmission and control and also affect the analysis and use of video post-processing. In recent years, flicker removal research for video surveillance becomes an international hot spot.This paper mainly studies the technology of flicker removal based on nonlinear model for video surveillance. It proposes different flicker removal methods to different kinds of flicker which may occur in video surveillance. The main structure of the thesis is as follows:(1) According to the prevalence of global flicker in video surveillance, this paper studies a flicker removal method based on nonlinear model. Firstly, the method introduces Hurter-driffield density versus log exposure characteristic which reduces a nonlinear relationship between the intensity and the corresponding deflicker. Secondly, a histogram is drawed by the intensity error between a reference and a degraded frame, choosing the intensity value which corresponds to the most frequence as the corresponding intensity error. Lastly, the profile is obtained from the intensity error using the least square polynomial fitting, which can remove the global flicker well.(2) According to the local flicker in video surveillance, it is difficult to remove the local flicker for the basic nonlinear model. This paper studies a block-based flicker removal method which image is divided into several blocks and uses the non-linear model in the every block.Taking into accout the brightness frequecy, the reliability weighting estimation which is presented in the paper makes flicker profile be more smoother. A local correction at this stage can introduce blocking artifacts due to neighboring blocks being compensated with different parameters. Boundary effects are avoided using weighted bilinear interpolation. The proposed method is a substantial improvement over our previous work.(3) According the flicker presents in the motion video, this scheme takes into account the movement of the scene and the camera during video shot. Motion compensation can be employed before flicker removal. This paper uses spatial segmentation to identify regions of homogeneous brightness, the nonlinear model in the every region which is based on the block to obtain reliable estimation of flicker parameters. The profile is obtained from the intensity error between the referecne and the degraded frame using the reliability weighting estimation. Our scheme incorporates an novel motion-compensated grey-level tracing approach in the deflicker compensation,which addressed problems arising from varying scene motion and illumination, which make a good compensation of long duration surveillance video.Experimental results show that the deflicked video have a big improvement in the visual and also have a progress in the mean intensity. All of the three methods can satisfy the requirements of flicker removal.