We present a biologically inspired spiking neural network which is able to detect contours in grey level images by synchronization of neurons. This network is made of integrate-and-fire neurons, spaced on a triangular network, whose oriented receptive field is constructed by a wavelet which specifically detects edges. The neurons are excitatorily and locally connected between receptive fields that tend to detect the same contour. A contour, if its width is not too large, activates a chain of neurons, with some heterogeneity in the inputs. The capacity of a chain tosync hronize with respect tosuc h heterogeneity is studied. Synchronization on a contour is found to be possible for a sufficiently large width.