The edge extraction from images plays a key role in the elementary processing of computer vision system, but it’s still a bottleneck problem. This thesis mainly paid attention to the issues related to the detection of the edges from the complicated nature background, and an improved method for edge detecting was presented based on the fractal features. Furthermore, a new method about how to calculate the fractal dimension and the intercept feature was proposed by applying the property of fractional Brownian motion to the algorithm proposed. Then, based on the fractal dimension and intercept features, a measure method for discriminating edges of man-made object from natural scenes was provided. Experiments show that the proposed algorithm can detect the edges effectively, at the same time the runtime is reduced greatly.