"Magic" Reconstruction: Compressed Sensing Options

[klonuo]

What do you guys, think about this?

When I first heard about compressed sensing, I was skeptical. There were claims that it reduced the amount of data required to represent signals and images by huge factors and then restored the originals exactly. I knew from the Nyquist-Shannon sampling theorem that this is impossible. But after learning more about compressed sensing, I’ve come to realize that, under the right conditions, both the claims and the theorem are true.

Compressed sensing promises, in theory, to reconstruct a signal or image from surprisingly few samples. Discovered just five years ago by Candès and Tao and by Donoho, the subject is a very active research area. Practical devices that implement the theory are just now being developed.

It is important to realize that compressed sensing can be done only by a compressing sensor, and that it requires new recording technology and file formats. The MP3 and JPEG files used by today’s audio systems and digital cameras are already compressed in such a way that exact reconstruction of the original signals and images is impossible. Some of the Web postings and magazine articles about compressed sensing fail to acknowledge this fact.

more...

[saratoga]

What do you guys, think about this?

Its a major area of academic research thats quite popular right now in imaging processing and military applications. Its not really useful for audio though since audio data rates are extremely low, so theres no reason to tolerate the huge increase in complexity associated with sampling on a sparse basis.

Its great for things like imaging though where you might want to do billions of measurements per second and only have a sensor capable of a fraction of that.