The integration of CMOS cameras with embedded processors and wireless communication devices has enabled the development of distributed wireless vision systems. Wireless Vision Sensor Networks (WVSNs), which consist of wirelessly connected embedded systems with vision and sensing capabilities, provide wide variety of application areas that have not been possible to realize with the wall-powered vision systems with wired links or scalar-data based wireless sensor networks. In this paper, the design of a middleware for a wireless vision sensor node is presented for the realization of WVSNs. The implemented wireless vision sensor node is tested through a simple vision application to study and analyze its capabilities, and determine the challenges in distributed vision applications through a wireless network of low-power embedded devices. The results of this paper highlight the practical concerns for the development of efficient image processing and communication solutions for WVSNs and emphasize the need for cross-layer solutions that unify these two so-far-independent research areas.