Performance of sub-pixel registration algorithms in digital image correlation

TLDR

Digital image correlation has become a popular tool for full‑field displacement and strain measurements, with sub‑pixel registration algorithms key to improving accuracy, yet quantitative comparisons of their performance are scarce. The study investigates the performance of three commonly used sub‑pixel displacement registration algorithms in terms of registration accuracy and computational efficiency. The comparison is performed on computer‑simulated speckle images. The iterative spatial domain cross‑correlation algorithm (Newton–Raphson) is the most accurate but also the slowest, and is recommended for these applications.

Abstract

Developments in digital image correlation in the last two decades have made it a popular and effective tool for full-field displacement and strain measurements in experimental mechanics. In digital image correlation, the use of the sub-pixel registration algorithm is regarded as the key technique to improve accuracy. Different types of sub-pixel registration algorithms have been developed. However, little quantitative research has been carried out to compare their performances. This paper investigates three types of the most commonly used sub-pixel displacement registration algorithms in terms of the registration accuracy and the computational efficiency using computer-simulated speckle images. A detailed examination of the performances of each algorithm reveals that the iterative spatial domain cross-correlation algorithm (Newton–Raphson method) is more accurate, but much slower than other algorithms, and is recommended for use in these applications.