The invention relates to a contour detection method based on visual mechanism dark edge enhancement. The method comprises the following steps: firstly, simulating retina photoreceptor cell response characteristics, and proposing a dark field adjustment model based on local brightness characteristics; then, in combination with the scale orientation characteristic of the classical receptive field ofthe ganglion cells, obtaining a primary contour response, and extracting global contour information by utilizing a principal component analysis method; then, when a retinal ganglion signal is transmitted to an outer knee body, simulating a side inhibition effect of a non-classical receptive field, further introducing a nerve cell sparse response characteristic, and cooperatively inhibiting a background strong texture of a primary contour; simulating the enhancement effect of micro-motion information on contour perception understanding so that background weak textures are reduced, and then using adaptive dynamic synapses for outputting and transmitting external knee body impulse responses to a primary visual cortex; and finally, correcting the primary contour response through global contour information, and outputting and quickly fusing the corrected primary contour response with the primary visual cortex response to generate a more accurate and effective contour detection result.