Design of a DSP-based CMOS imaging system for embedded computer vision

With the advance of image processing techniques and the fast reduction of image sensor costs, embedded computer vision is becoming a reality. This paper presents the design and implementation of an integrated CMOS imaging system based on DM642. The system can acquire VGA resolution image at 100 frame/s. The on-board 10 Mbps Ethernet connection and the UART as well as digital I/O ports provide convenience for direct interface to other intelligent devices. The on-board software provides image acquisition function, network-based control, network-based data transfer, JPEG compression, image pre-processing functions, edge detection function. Automatic exposure control and automatic white balance are realized in the system. Experiments show that the CMOS imaging system has a good performance in terms of imaging quality, with much potential for being used as intelligent vision system.