When we talk about digital cameras, we usually think in terms of megapixels—millions of pixels—but vision systems tend to work at a much lower resolution. Because of the processing power involved (and the difficulty in upgrading), it’s not unusual to see cameras operating at a 640×480 resolution in an industrial environment. Perhaps in some applications though, an even lower resolution could be useful, like the 30×30 SilicanRIK.
This camera board can be programmed in C and C++. It features a frame rate of up to 105 FPS, though processing speed can be much slower depending on the program running. It features 9 digital I/O pins and 3 analog inputs, so perhaps this might be a good option to look into for an embedded design.
Although a 9 ‘centipixel’ camera doesn’t sound very impressive, these cameras have been used for face detection, vehicle classification, and speed measurement. Seen below in a video featuring the older SilicamIGO, two low-resolution cameras can be used together for stereoscopic depth estimation. The resulting visual certainly looks strange on YouTube, but I imagine it would be even weirder in person.