TOP-DOWN LEARNING OF LOW-LEVEL VISION TASKS

Perceptual tasks such as edge detection, image segmentation, lightness computation and estimation of three-dimensional structure are conside red to be low-level or mid-level vision problems and are traditionally approached in a bottom-up, generic and hard-wired way, An alternative to this would be to take a top-down, object-class-specific and exampl e based approach. In this paper, we present a simple computational mod el implementing the latter approach, The results generated by our mode l when tested on edge detection and view-prediction tasks for three-di mensional objects are consistent with human perceptual expectations, T he model's performance is highly tolerant to the problems of sensor no ise and incomplete input image information, Results obtained with conv entional bottom-up strategies show much less immunity to these problem s, We interpret the encouraging performance of our computational model as evidence in support of the hypothesis that the human visual system may learn to perform supposedly low-level perceptual tasks in a top-d own fashion.