Embedded computer vision framework on a multimedia processor

In this paper we present a framework for implementing computer vision algorithms on embedded processors using the analog devices Blackfin BF561 processor. This framework is designed to consist of a set of simple functions that allow complex tasks to be performed, such as object tracking or object recognition. Special consideration is given to the methods of optimization that are needed to achieve real-time performance. Given the low cost and low power consumption of embedded devices, the applications for this framework are very widespread and have the potential to be seen in many everyday devices. The completed framework is capable of processing algorithms at 15 frames per second with 24 bit colour, 360 times 240 pixel images which supports this approach as a useful direction for low cost, low power computer vision platforms.