Edges of objects often provide important features in pattern recognition application. How to extract the edge effectively in noise image is a difficult problem in the field of image analysis. Heuristic search algorithm is often used to extract edge

but because this method utilizes the fixed start points

fixed guide measurement and processes one off search

it is often very sensitive to noise. In this paper

a stochastic heuristic search method is presented. It firstly uses repetitive random searches to obtain various possible independent edge trajectories

then self reinforces and accumulates the search trajectories respectively

at last

extracts the edges relying on the result of the accumulation of self reinforcement. Lots of experiments show that our method can extract edges effectively and suppress the noise at the same time. Comparing with heuristic search algorithm

we find that our method is superior.