A Configurable Window-Based Processing Element for Image Processing on Smart Cameras

Image processing in real-time is known as a high processing-power consuming task. Smart cameras deal with this problem by adding embedded processing resources to the camera architecture. This way, early low-level vision tasks can be done internally, before transmission to the host system and helping to prevent from communication bottlenecks. This paper presents a configurable window-based processing element, tailored for implementation on a FPGA device. Window-based operations are often required for image filtering, feature detection, tracking, etc. In this work, some particular points of these applications are analyzed, and it is shown that a general computational model can be extracted. This model is applied to conceive a processing element, able to perform several window-based operations. A hardware implementation of the configurable window-based operator is proposed, and a smart camera platform, able to host such processing element, is presented and detailed.