The vast majority of car-following models are lack of the consideration of human drivers' characteristics. Based on the fact that each driver of a following vehicle perceives closing-in or shying-away a leading vehicle in front of him/her, primarily due to changes in the apparent size of the leading vehicle, we improved the full velocity difference (FVD) model and presented a visual angle car-following model. This model is in view of the stimulus-response framework and uses the visual angle and the change rate of the visual angle as stimulus. Results from linear analysis showed that the neutral stability line is asymmetry and the width of the leading vehicle has a great impact on the stability of traffic flow. Numerical simulations obtained the same results as theoretical analysis clearly such as density wave, shrinking hysteresis, asymmetry and wide scattering. Thus, the introducing of the visual angle can explain some complex nature of traffic flow and contribute to the design of more realistic car-following models.