Through the mechanism called lateral inhibitory existing among the visual neurons, visual information can be selectively detected. Simulating this mechanism among visual cellars through a mathematical method, a shunting inhibitory retinal cellular neural networks(SIRCNN) model based on early vision is provided. The stability analysis of its dynamics and the equivalent circuit diagram are presented and a method of detecting edges based on this model is provided. From the test result, such a method based on the biological visual mechanisms can efficiently enhance and detect edges . Because the cellular neural networks are uniquely suitable for the high-speed parallel computation and easy to implement in hardware, this model has more potential in real-time image processing.