Progressive Background Image Generation of Surveillance Traffic Videos Based on a Temporal Histogram Ruled by a Reward/Penalty Function

In most surveillance traffic systems the background image is used for segmenting the moving objects. In this paper, we propose a method for progressive and constant renewal of the background image of a video sequence. Our method is divided in three parts: frame difference to get an actual information of the video sequence, reward and penalty function that modifies a temporal histogram, and the background generation uses the information found in the histogram. The principal characteristic of the method is the constant regeneration of the histogram using neighborhood information of intensity levels, penalizing the occurrence of foreign objects and rewarding the values of static objects. For that, the histogram values are modified regionally and its variation correspond to a normal distribution. The principal advantages of our proposed method lies in the progressive increment of the background pixel intensity occurrences, the fast reaction to changes between foreground and background segments and the simplicity of the method that can be used in real time applications. At the end of this paper, experimental results demonstrated the feasibility of the proposed technique in different weather conditions and traffic situations.