Compressive imaging

Compressive sensing may be utilized to reduce the number of sensors in digital cameras. In conventional cameras, the sampling and compression stages are separated which form a non-efficient pair since most of the sampled data is disposed of during the compression process. Compressive sensing reduces such data acquisition inefficiency by combining the sampling and compression stages into one compressive sampling stage.

Existing linear image transforms encourage the utility of compressive sensing for sensing naturally correlated image signals. Furthermore, over-complete and trained dictionaries allow even more sparse representations of images and color images in general.

Direct application of compressive sensing to digital imaging involves changing the hardware of the camera: a compressive sample which corresponds to one row of a dense sampling matrix is a linear combination of different pixels within the image. A camera structure for such sampling has been proposed and fabricated by digital signal processing group in RICE.