Image analysis through the Wigner distribution function.

The Wigner distribution (WD) for discrete images is computed and applied for texture analysis. First, the WD for different test images with spatial frequency and gray level information content is computed, and the most important features of the WD are outlined. Likewise, the original image with 1-D variation is recovered from the discrete WD; local filtering operations have been performed on the distribution, and a filtered version of the original image is then obtained. Since the WD entails simultaneous representation of the intensity distribution and of the spatial interdependence, it seems specially adequate for processing of textured information. In this way, texture recognition is performed through the extraction of features from the WD, and the results are compared with other methods. Finally, to avoid the great computational effort that requires this kind of representation, a hybrid optical-digital processing system is proposed for the texture recognition process.