Conversion gain is an important parameter of low-light CMOS image sensors, and realizing improvements in conversion gain helps to enhance their signal-to-noise ratio, which in turn improves their sensitivity and imaging quality. Because the conversion gain is inversely related to the parasitic capacitance of a pixel, this paper proposes a two-dimensional physical model of the parasitic capacitance of a 4-transistor active pixel based on the self-alignment technique to effectively improve the conversion gain. The proposed model establishes the relationship between pixel parasitic capacitance and injection conditions, which include injection dose, injection energy, and reset voltage. The calculated results of the model are in good agreement with the TCAD simulation results, and the variance between them is less than 0.0028 fF², verifying its accuracy. The proposed model can be applied to the design and optimization of high-performance image sensors, especially high-sensitivity low-light image sensors.