JPEG-optimized image sensor

Posted December 19, 2005

This article describes a cute little trick for saving power in image sensors. Folks realized that most of the time, the raw pixel data from a picture was just immediately getting handed to a microprocessor to compress it, usually into JPEG format. So not only was some of the image data getting always discarded, but extra processing power was being burned doing the compression itself.

So instead, they decided to arrange the pixels in the sensor into a different spatial pattern than just a rectangular grid. The pixels end up being aligned with the points in the discrete cosine transform (used in JPEG compression) of maximal weight, so you only gather the data that's most useful, and furthermore your compression algorithm is much easier since you don't need to perform a lot of weight multiplications, saving power.

The catch is that your sensor isn't quite as efficient in gathering light, since the pixels aren't packed solid, but for cheap sensors that can be a reasonable tradeoff. So, a nifty idea in itself, and also a good technique to think about in other situations: how can I optimize my sensor to save myself work farther along in the analysis chain?