To eliminate the effect of tilt-shift error on the accuracy of phase-shifting interferometry (PSI), a fast and accurate tilt-shift-immune phase-shifting algorithm based on the self-adaptive selection of interferogram subblocks and principal component analysis (SSPCA) is proposed. First, each interferogram is divided into several subblocks, and principal component analysis and the least-squares method (LSM) are applied to obtain the phase-shift value of each subblock. Next, according to the correlation coefficients between each phase-shift curve, valid and invalid subblocks can be distinguished. Finally, all phase-shift values of the valid subblocks are used to fit the tilt phase-shift plane, and phase results can be obtained using the LSM. Simulations indicate that the accuracy of SSPCA can reach 0.03 rad both for small (1 rad) and large (${2}\pi $2π rad) tilt amplitudes, and it takes only one-tenth or less of the processing time of iterative algorithms. Experiments proved that SSPCA can be applied even without a precision phase shifter and thus provides a low-cost approach for PSI with both high precision and speed.