Processing primitives for image simplification

We have already seen in Chap. 2 that the simplification of the scene is a key issue when it comes to save energy in vision-enabled WSN nodes. It permits to reduce the clock speed of the digital processor as well as the number of memory accesses, two important parameters concerning the power consumption of the nodes. Obviously, the energy cost of obtaining a simplified version of the scene and processing it must be lower than the energy cost of directly processing the raw representation of the scene. Therefore, it does not suffice to define image processing primitives capable of achieving such simplification. Their hardware implementation also plays a crucial role in order to reach a really high degree of energy efficiency.