Human Perception

Introduction to human perception “The goal of computer graphics is not to control light, but to control our perception of light. Light is merely a carrier of the information we gather by perception.” (Jack Tumblin, James A. Ferwerda) Outputs of computer graphics are intended to be observed by human subjects. As human vision has several limitations, the knowledge of the human visual system (HVS) and of the human perception can be utilized to improve the performance of various computer graphics algorithms. In the field of computer graphics the knowledge of the human visual system usually takes the form of the computational models of human vision. Such a model can be incorporated at various areas of computer graphics. 1 One of the areas where the incorporation of human vision models is extremely beneficial is the image quality assessment and the image comparison. Image quality assessment and comparison metrics play an important role in various computer graphics applications. They can be used to monitor image quality for quality control systems, they can be employed to benchmark image processing algorithms, and they can be embedded into an image processing system to optimize the algorithms and the parameter settings. It is well known [49], that classical comparison metrics like Root Mean Square (RMS) error are not sufficient when applied to the comparison of images, because they poorly predict the differences between the images as perceived by the human observer. To solve the problem properly the visual differences predictors have evolved. The main part of visual differences predictors is typically a model of early vision, so that they perform well when comparing visually very near images. However their performance when comparing quite different images with respect to the contained information is poor. The predictor capable to incorporate such a behaviour would be valuable in the image database retrievals, to evaluation of the perceptual impact