Deblurring images using projection pursuit learning network

The problems of image deblurring and noise reduction are addressed. We present a system that require little or no prior knowledge of the blurring (noise) source. Projection pursuit learning network (PPLN) is used to achieve this goal. We show that a PPLN trained with an image A that was blurred using Gaussian produces promising result when tested on an image B where the blurring source is not Gaussian. Similar arguments can be made in the case of noisy images. The experimental results presented point to the fact that the trained PPLN can successfully handle blurred as well as noisy images