The temporal phase shifting technique is a powerful tool that is used in numerous industrial, research and development applications, such as range acquisition, industrial inspection, reverse engineering, and object recognition. However, the accuracy of temporal phase shifting technique is limited by the phase-shifting uncertainties resulting from the imperfect conditions in the real testing environments, such as the miscalibration. So, there is a strict requirement on the precision of the phase shifter. To remove this requirement, there have been reported works about phase reconstruction with only two frames. Two-shot phase reconstruction, i.e., the recovery of a complex-valued signal from two intensity patterns that are nearly identical apart from a random and unknown phase shift between them, which can reconstruct the phase without local sign ambiguity using the minimum number of interferograms with higher measurement accuracy than the single-frame algorithm, is especially prevalent in imperfect testing conditions where requires reducing the detrimental effects on the measurements from mechanical vibrations, ambient air turbulence, and temperature changes. In the implementation of the technique, the filtering procedure of the fringe normalization, background removal, or noise reduction must be applied to the two arbitrarily phase shifted patterns before the phase demodulation. However, due to certain experimental imperfections of any physical system that are hard to model accurately, we cannot assume any particular behavior for the intensity pattern in a general case. In this case, the two-shot phase reconstruction technique may have an unsatisfactory filtering effect during the phase demodulation. Recently, much special attention is paid to the algorithm that does not require pre-filtering. In this work, we proposed a novel two-shot phase reconstruction method that can bypass the pre-filtering with the help of the least squares iterative phase shifting algorithm. In our method, a rough estimation of the phase shift is obtained by correlation operation of two background removed fringe patterns under the approximate orthogonal characteristic of a trigonometric function. Thus, we can compute the phase using two frame phase shifting algorithm. Then, using this initial phase shift estimation, both the phase shift and background light improved by solving other N linear equation systems in a frame-wise manner. Further, the two background removed fringe patterns are updated. Iterating this process, it eventually settles down in the searched phase. In essence, this method is a two-frame version of the AIA algorithm except for the use of background terms, which is usuallydiscarded in the standard AIA algorithm. To reduce the effect of spatially varying background intensity and modulation amplitude, we divide the captured interferogram space into several blocks (for example, 150×150). if the blocks are sufficiently small, we may consider that the background intensity and modulation amplitude in each block do not have pixel-to-pixel variation and can be assumed as constants. Then we select one block, and the necessary parameters are estimated by the phases and the intensity in this block. Our novel approach consists of the following steps: 1) Estimate an initial phase step by s correlation operation of two background removed fringe patterns, 2) Compute the measurement phase using the initial phase step, and 3) Use the obtained intermediate parameters for continually updating both the two filtered fringe patterns and the phase step until the accurate phase is reconstructed through the searched phase shift. To demonstrate, the proposed method is compared with other state-of-the-art two-frame random phase-shifting algorithms: CV and PCA&LSI, for they can accomplish the phase reconstruction universally and accurately. We choose the typical scenarios, namely open-fringe pattern, and closed-fringe pattern as our test case. Both simulations and measurement experiment results obtained using the proposed method indicate that it is a simple, albeit robust solution for phase extraction from two-frame unknown phase shift fringe pattern.