Lessons from Human Vision

In this concluding chapter, it will be useful to compare some of the image analysis capabilities that have been described to the performance of the human visual system. Several times throughout this text, reference has been made to the methods that appear (based on physiological tests) to be used by humans in dealing with visual information (and indeed other animals - see for example Lettvin et al., 1959; Hubel & Wiesel, 1962). These insights have provided valuable guidelines to methods that are robust, fail gracefully in difficult situations, are usually extremely economical in their computational needs, and above all, work. Examples include stereo pair fusion, feature recognition and identification, correction for shadowing effects, and the location of feature edges even when they are incomplete or obscured in noise.