Simultaneous estimation of multiple order phase derivatives using deep learning method in digital holographic interferometry

For non-contact deformation testing, digital holographic interferometry is a prominent optical technique where the first and second order interference phase derivatives directly embed information about the strain and curvature distributions of a deformed object. Hence, reliable extraction of multiple order phase derivatives is of great practical significance; however, this problem is marred by several challenges such as the need of multiple differentiation operations, complex shearing operations and performance degradation due to noise. In this paper, we introduce a deep learning approach for the direct and simultaneous estimation of first and second order phase derivatives in digital holographic interferometry. Our method's performance is demonstrated via rigorous numerical simulations exhibiting wide range of additive white Gaussian noise and speckle noise. Moreover, we substantiate the practical efficacy of our proposed method for processing deformation fringes acquired via digital holographic interferometry.