EFFICIENT VIDEO CODING IN REGION PREDICTION IN ONLINE VIDEO SURVEILLANCE

摘要

There is a recent trend to implement video coding techniques for video surveillance. Various coding techniques had been proposed to enhance estimation accuracy in progressive video coding. As the conventional coding approaches have constraints to optimize image processing for moving bodies whilst the background is typically static. The problem magnifies in case of rotating cameras. The challenge in this case is to segregate the dynamic entities while the background rotates at a fixed pace. This paper presents an approach for the improvement of error free coding in video surveillance by employing rotating cameras. A least mean estimator approach is used and the recurrent full search motion estimator logic is defined for the prediction of foreground moving elements from the video sequence, which comes from a rotating sensor. The benefits that we derive from this method are more accurate detection of the actual moving object, thereby, reducing data redundancy by eliminating the nonessential background information.