"Achieving real-time performance for image processing and computer vision applications is becoming
increasingly crucial because of the growing complexity of computer vision algorithms and inefficient
exploitation of hardware’s processing capabilities. To achieve this goal, we need to efficiently map the optimization
opportunities inherent in these algorithms to the kind of architectural optimizations that are capable
of fully utilizing available hardware resources. A survey of the existing architectures will be presented. Also, a framework, developed for the analysis of power savings obtained by these optimizations, will be presented."