A multiple camera surveillance system is typical example of a Distributed Video (DV) system with many-to-one topology which demands the use of new paradigm with multiple encoders, installed at number of locations and very few decoders in a control room. This paradigm open- up new frontiers for the research community in designing low cost encoders even at the cost of expensive decoders. A solution to this problem based on information theory finding of 70s, by Slepian and Wolf , for lossless encoding, and followed by the work of Wyner and Ziv , for lossy encoding. In the last few years there has been significant research activity in the design and implementation of video codec based on these findings. Our research work is also part of that effort. In this paper we will present a modified Wyner-Ziv codec, that take advantage of slow motion activity, which is typically of surveillance data. GOP is selected dynamically by accessing motion activity and also puncturing bit rate varies adaptively. Also the proposed architecture takes advantage of intra-frame coding for both key frame and Wyner-Ziv frames.