Image recognition in the presence of laser speckle

Two techniques for performing digital image recognition on speckled images are presented. The speckle is treated as a source of randomness that is used to perform Monte Carlo calculations of the correlation between an input scene and a reference scene. In the first technique, correlation is determined from intensity measurements of the speckled image. In the second, the input image is clipped, and correlation is determined from the locations where the speckle intensity is above the clipping threshold. Several correlation filters are applied to the speckle and clipped-speckle techniques. The filters are compared according to the ability to distinguish a set of images. The experiments were performed using coherently illuminated rough objects.