Fringe projection profilometry (FPP) has been widely used in three-dimensional (3D) profile measurements for its efficiency and high accuracy. As an offshoot of FPP, color fringe projection profilometry (CFPP) has the advantage of being dynamic, non-contact, and full-field, which makes it one of the most valuable 3D profile measurement techniques. However, phase errors may arise in CFPP when the phase is retrieved by phase-shifting algorithms. This is due to the gamma nonlinear response, color coupling, and color imbalance of electro-optical devices. In this paper, a multi-confusion matrices (MCM) correction technique is proposed to reduce the phase error by compensating the captured color fringe pattern intensity. This can be accomplished in three steps. Firstly, project 30 ~220 (with the step of ten gray levels) uniform intensity patterns on a calibration plate, and then extract the intensity information to construct multi-confusion matrices (MCM) and LUTs (look up tables). Secondly, search optimal confusion matrix (OCM) from MCM to correct the fringe pattern. Finally, retrive the unwrapping phase using the corrected fringe pattern. The simulation and experimental results indicate that the proposed approach can reduce phase errors and use one color fringe pattern to reconstruct 3D topography.