The performance of the human visual system in extracting noisy figures from a noisy background is astonishingly good. Even in situations of very poor contrast, boundaries emerge clearly. Conversely, typical edge detectors fail to give good results for such images. In an attempt to explain the discrepancies in these performances we have developed a neural network model relying on two assumptions, both of which are based on neurophysiological findings. Firstly, the processing of visual information is considered to be asynchronous: stimuli are delayed accordingly to their intensity. Secondly, emergent boundaries have the property of producing coherent responses corresponding to the (near)-simultaneous responses of cells in the cortical orientation columns. Results show that such neural network can indeed benefit from the asynchrony when treating images with ratio signal-to-noise particularly low.