Application to multispectral images of a blind identification system for blur, additive, multiplicative and impulse noises

We address the problem of the identification of the preponderant degradation affecting an image, in the context of blind, processing where the identification has to be made from the observed image. Considering that the main difficulty for any pre-processing treatment is to find the good balance between the two contradictory aspects of preserving the fine details and removing the degradation effects, the estimation from the observed image of the degradation characteristics is crucial to choose the algorithms and the order in which to apply them. The degradations considered here involve a defocusing blur and a noise or a combinations of both. The noise can be additive, multiplicative or impulse. The system presented here is tested on images obtained from the CASI airborne hyperspectral imaging sensor.