Cameras, whether a still camera based on film or an electronic camcorder, are essentially passive in that they only record whatever image is presented to the lens. But if we could insert advanced image processing functions between the image sensor and the storage medium, it would be possible to create a more dynamic smart camera. An obvious application is fixed, low-cost cameras that are used in security applications.
Basically, such cameras passively record a room or public area where often nothing is happening. Security systems rely on vigilant personnel to spring to action when some activity occurs. It would be far better to have a camera that automatically responds to action, sending an alert when activity is detected and also recording the relevant action for storage or transmission over a network.
But to implement such a design, a low-cost, flexible video processor with a large capacity to crunch image data sets is required. We believe that Oxford Micro Devices Inc.'s Ax36 video DSP chips provide the combination of processing capability and low-power operation that smart cameras would need. For example, the processors include an on-board motion-detection and estimation coprocessor that can quickly identify any changes in a video stream.
Another important piece of the smart-camera architecture is advanced image-processing algorithms. Much has been learned in the past 10 years about image manipulation. For example, it is now possible to change the depth of field, focus, and f-stop of a physically fixed camera entirely through image processing. In addition, the angle of view can be changed. These new image-processing capabilities become really significant when combined with networking, allowing users to remotely control the characteristics of a camera simply by downloading software.
A consumer application opportunity that this capability provides is the transparent update of an existing closed-circuit video system. Smart cameras could replace existing units, and the only physical addition to the video network, beyond the new cameras, would be a phone line. The cameras could still be monitored and controlled by the existing network, but they could also be controlled via the Internet, so that the closed system could be opened to a wider corporate or government monitoring organization.
With sophisticated software in solid-state memory, it would be easy to add features for a variety of different applications. It becomes possible to build a family of products using the same image processing motherboard, changing the functionality easily, both at the time of manufacture and later via the Internet.
Using the Ax36's flexible image-compression capability, still cameras could produce images with very high spatial and color resolution, and frame rates at least as fast as a motorized film advance could be introduce on a still camera. This requires that individual images be compressed very rapidly so that the speed of compression does not limit the frame rate. Low-cost digital camcorders could produce broadcast-quality images, if not images with even higher resolution as HDTV nears. Web cameras that use the latest, constantly evolving video-compression algorithms could be optimized to communicate and store the best images possible with the least bandwidth This scenario is based on a very high-throughput, parallel processing capability optimized for image data sets. But equally important in the consumer arena is the need to provide an intermediate level of processing capability at very low power. The winning digital still camera designs in today's market offer long battery life.