Bayesian approach to reconstruction of textural image parameters

There are three main tasks in object decipherment. These tasks are object segmentation, type recognition and parameter reconstruction. The main drawback of these approaches are low recognition rate, problems with the recovering of complex parameters such as species components and the requirement for a vast number of training samples or prototypes. We propose a novel method, based on a combination of the abovementioned techniques, using a small number of the prototypes and Bayesian interpolation between the prototypes. A special technique allows effective mapping of the reconstruction results.