Visual surveillance with a distributed ad-hoc sensor network will enable reliable monitoring and control of a variety of applications that range from civil to military. However, due to the large volume of multimedia data associated with visual sensors, the surveillance scheme has to be architected with a goal of energy conservation and extending life expectancy. In this paper, we briefly discuss the computation and communication energy costs associated with a sensor node and explore the possibilities of extending battery life by choosing to perform local processing. We propose a Discrete Wavelet Transform (DWT) based multi-resolution surveillance schema designed to reduce power requirement. We then compare and contrast the computational and communication power expended by a traditionally used JPEG-based scheme and illustrate the significant power savings achieved by the proposed surveillance system.