Extracting Background Model in Video Surveillance by Using Hybrid Techniques

One of the most important methods of video surveillance is background subtraction and one of its main steps is modeling of initial background image. In this paper the initial background image is formed by using hybrid techniques (i.e. k-mean and histogram). When the proposed method is employed, the outcomes proposes that the correct color value of highest frequency in the greater cluster leads to modeling of the initial background image, which in turn results in more accurate image of background . This accuracy is measured through PSNR (Peak Signal to Noise Ratio) metric. Furthermore, the current study proposed a comparative study with other four methods (mean, median, histogram and k-mean) through the use of PSNR metric.