To expand the reliability of interferometry technology, this paper proposes a random two-frame algorithm with high accuracy, high robustness, and immunity to tilt phase-shift. This method uses the equivalence of inter-frame phase-shift and intra-frame phase difference to mine light intensity pixels carrying new phase-shift from different images. Then, the linear random phase-shift plane is fitted by least squares, and the inverse tangent relationship is used to obtain a high-precision phase distribution. This technology uses the principle of light intensity equivalence to fit the linear phase-shift plane and does not require any iterative process. It can effectively suppress the influence of tilt phase-shift while ensuring computational efficiency. The paper verifies that the proposed algorithm has excellent performance in both tilted and non-tilted conditions through simulation and experimental comparison.