We are witnessing a major shift in the way computer vision applications are implemented, even developed. The most obvious manifestation of this shift is in the platforms that computer vision algorithms are running on: from powerful workstations to embedded processors. As is often the case, this shift came about at the intersection of enabling technologies and market needs. In turn, a new discipline has emerged within the imaging/vision community to deal with the new challenges:

embedded computer vision (ECV).

Building on synergistic advances over the past decades in computer vision algorithms, embedded processing architectures, integrated circuit technology, and electronic system design methodologies, ECV techniques are increasingly being deployed in a wide variety of important applications. They include high volume, cost-centric consumer applications, as well as accuracy- and performance-centric, mission-critical systems. For example, in the multi-billion dollar computer and video gaming industry, the Sony EyeToyTMcamera, which includes processing to detect color and motion, is reaching out to gamers to play without any other interfaces. Very soon, new camera-based games will detect body gestures based on movements of the hands, arms, and legs, to enhance the user experience. These games are built upon computer vision research on articulated body pose estimation and other kinds of motion capture analysis. As a prominent example outside of the gaming industry, the rapidly expanding medical imaging industry makes extensive use of ECV techniques to improve the accuracy of medical diagnoses, and to greatly reduce the side effects of surgical and diagnostic procedures. Furthermore, ECV techniques can help address some of society’s basic needs for safety and security. They are well suited for automated surveillance applications, which help to protect against malicious or otherwise unwanted intruders and activities, as well as for automotive safety applications, which aim to assist the driver and improve road safety.