Abstarct: The video devices usually produce huge data. There is a challenge for saving or transporting this massive data. The video compression systems are devised to figure out this problem. These systems use interfrxame redundancy to compress this enormous data. For detecting the redundancy, frxames are divided to some same size blocks afterward block-matching component of compression system examines these blocks, to avoid from saving repeated block. The block-matching component is very time consuming component in the video compression system. In addition, if this component does not operate properly subsequently compression percent would be reduced. However, in common video compression systems such as MPEG2, explore all the bocks as a result these systems need enormous computation. In these thesis by using background modeling, initially active regions are identified, the active region means a region of frxame which moving objects are placed in it, after that we limit the block-matching operation to these regions. To detect active regions, in first step, a model of background is identified and noises are removed, then output of previous stage submits to another component. In this component, pixels of the frxame are classified to three categories: object, shadow and background class. Finally by employment these information, the active regions are detected and the block-matching operations are limited to these regions. The outputs of proposed approach are judged against of the four methods (2D- logarithm search، UCBD،4SS and CSD). The results demonstrate the new optimal system, is very faster than another methods. Hence, the experiments express volume of output of new method is half of volume of output of non-optimal systems.