Large numbers of sensors networked together form a powerful infrastructure for a wide variety of applications in health, military, home, manufacturing, and disaster areas. Networked video sensors over a geographical area can automatically detect the objects in that geographical area. However, real-time central processing of the large amount of data generated by the individual image sensors places significant demands on the bandwidth and the central processor. In this paper, we address this issue by introducing a novel concept of super-sensor that is based on self-organization and collaboration between several tens of sensors. As an example, we describe a histogram calculation that uses recursive doubling for global collaboration between sensors. We compare the performance of our networked super-sensor with conventional image processing algorithms run on a single processor.