This document investigates key issues in extracting information from images. Perimeters of objects are key features in human recognition, and are found through edge detection. Several edge detection methods are investigated in this paper, including fuzzy edge detection. Hough lines were drawn on the edges making use of 'Harris corner detection' to estimate the number of lines to draw. The lines were connected up into triangles and this was found to segment key parts of the images. The overall texture contained within a set images was analyzed, with it features being reduced by canonical variants. Classical classifiers and self organizing maps were used to analyze the textures, and showed very similar confusion.
展开▼