Abstract: The capacity of RF communication can usually be calculated using the Shannon formula, while the capacity of visible light communication (visible light communication,VLC) is different from that of RF due to the nonnegativity of the signal, and needs to be re-deduced to adapt to the characteristics of VLC. Firstly, considering the additive Gaussian white noise and the input signal correlation noise, the lower bound of the channel capacity in three scenarios is derived by using the method that the output entropy of the channel is always greater than the input entropy. Secondly, in order to solve the problem that the channel capacity of indoor VLC decreases significantly in the corner position, it is proposed to optimize the channel performance by tilting the receiver plane, and an optimization model is constructed to effectively improve the channel capacity. Finally, numerical simulations verify the accuracy of the lower bound of the deduced channel capacity, and analyze the distribution characteristics of indoor VLC channel capacity. The simulation results show that the larger the ratio \rho of the input-related noise variance to the input-independent noise variance, the smaller the channel capacity. When the room size is 6 m×6 m×4 m, the LED is in the center of the room, and the receiver is also in the center of the room, the channel capacity is the best, and the further the corner position, the worse the channel capacity. And when the receiver is tilted, the maximum inclination angle in the room is 60.6°, and the capacity after tilting is 0.68 bits/s/Hz, higher than the capacity before inclination. Input correlation noise has a large impact on VLC capacity, suggesting the necessity of considering input correlation noise when deriving channel capacity. At the same time, the comparison of the capacity before and after the receiver tilt shows that the channel capacity can be significantly improved by tilting the receiver plane appropriately.