Propagation function for accurate initialization and efficiency enhancement of digital image correlation

Digital image correlation (DIC) matches corresponding locations in two images by optimization correlation of the intensity of related pixels. Iterative algorithm is proved to be the most effective method to solve the optimization. However, it requires accurate initial guess of the unknown parameter. Existing methods directly use the deformation parameter of the current point as the initial for its adjacent points. We illustrate in this paper that direct transfer of parameter does not produce accurate initialization when deformation is present in the subset area of the current point. Instead, we propose an analytic propagation function which relates the parameters of adjacent points. Closed-form expressions of the propagation function are derived for both widely used first-order and second-order shape function, and they are validated using synthetic speckle images. Comparison with existing method shows that the propagation function provides more accurate initial estimate of the desired parameters and requires fewer iterations for the optimization to converge. Tests on simulated and real-world experiments also show that the propagation function can deal with large deformation, allow a wide range of step sizes, and reduce computational cost. (C) 2012 Elsevier Ltd. All rights reserved.