Fringe projection profilometry is a non-contact and highly efficient 3D measurement technique widely used in various applications. However, the nonlinear intensity response of digital projectors affects measurement accuracy. While increasing the number of fringe projections can reduce the errors caused by nonlinear problems, it significantly prolongs the measurement time. In order to improve both accuracy and speed simultaneously, a nonlinear phase error correction method based on multi-grayscale coding is proposed. The intensity response curve of the system is fitted by the grayscale images, and then the grayscale values of the phase-shifting fringe images are corrected to reduce the nonlinear error. In order to reduce the number of fringe projections and speed up the measurement, the multi-grayscale coding method is used to divide the phase interval by the order of the gray values of the same pixel in multiple grayscale images. The experimental results validate the efficacy of the proposed multi-grayscale coding method. An accurate phase calculation is achieved, and a single reconstruction can be achieved with only seven photos. After the nonlinear correction, the phase accuracy of the three-step phase-shifting algorithm is increased by 50.77%, and the reconstruction accuracy of the standard ball is increased by 46.38%.