A new approach in image analysis is presented, which makes explicit use of the notion of time. The purpose is to extract image features, such as edges, with the constraint that pertinent features should appear first. In particular, it is desired that strong contrasts are detected before weaker ones (contrast strength selectivity), or high curvatures are detected before straight lines (geometric selectivity). A simple electrical circuit is studied first to model contrast selectivity; a relationship between contrast strength and latency can be derived. The relation between stimulus geometry and latency is then explored by means of two types of receptive fields, namely lateral geniculate nucleus (LGN)-Iike and end-stopped cells. Concrete experimental results are presented that sustain our belief that time favors asynchronous processing in order to prevent overflow in the visual system by the visual information.