In this paper, we extend a lattice gas model recently proposed by Li et al, which considers the view field of pedestrian. An improved lattice gas model takes into account the effect of pedestrians' walking preference feature of empty area in the view field to simulate traffic dynamics of pedestrian counter flow. Three dynamic evolution processes under different pedestrian density are reproduced. The randomness of lane formation for different pedestrian density is found, and the probability of lane formation is given. Numerical simulations of relationship diagrams between the probability of lane formation and parameters of the system geometry size, the probability and the proportion of right walker flow, the probability and the strength of the drift, also the probability and the view field size are investigated. Results show that the extended model cannot form for the lane formation under a low pedestrian density, which is associated with the real pedestrian traffic. It is found that the density of pedestrian counter flow could be divided into 5 intervals, and there are differences in the dynamic evolution processes between these 5 intervals. This model and its result is useful for the study of the dynamic evolution process, and is helpful for raising efficiency of pedestrian counter flow in the channel.