This work proposes a point-specific self-calibration method to characterize film thickness distribution by exploiting the multiple detection capability of a home-built full-field ellipsometer. The self-calibration method offers a feasible route for retrieving calibration information from the actual real-time sample measurement in conjunction with the ellipsometric parameters, thus leading to error-free data after the elimination of systematic errors and addressing the problem of high time-consumption. With the help of the multiple detection capability of a full-field ellipsometer, we can further implement self-calibration for every point-specific pixel, termed as point-specific self-calibration to achieve a high-accuracy film thickness profile. The synthetic thickness distribution composed of structural-anisotropy pixels with tilted surface is utilized to demonstrate the potential of the proposed approach by retrieving the ellipsometric angles and the calibration parameters of every single pixel. A three orders-of-magnitude improvement in the accuracy of thickness determination was achieved in the simulation. To demonstrate the feasibility of the proposed approach, a film deposited on the Si substrate is measured in this work. This approach could be easily extended to implement thickness distribution measurements accurately and rapidly in other rotating-element ellipsometer cases.