Monday, October 25, 2004

Blind Deconvolution

After the obligatory speeches of the plenary session, Tony Chan of UCLA talked about blind deconvolution. It's called "blind" deconvolution because you don't necessarily know the point spread function convolved into your image, you still want to recover the original image. Tony talked about using total variation (TV) regularisation technique. Results looks fairly impressive.

He also talked about ways to distinguish the correct from alternative non-correct deconvolution solutions. For instance by using the "sharpness" of the recovered image.