Optimizing Metrological Efficiency: Comparative Analysis of Filtering Methods for 2D DIC

Digital Images Correlation (DIC) measures are widely used in order to calibrate models. It is very common for researchers to use filters on these measurements to improve the measurement resolution (decrease the noise level) before using them with various methodologies. However, the effects of these filters on the tradeoff between spatial resolution and measurement resolution are not fully understood. The objective of the study is to compare seven common denoising/filtering methods for 2D DIC with affine subset shape function, including image filters (Gaussian, Median), frequency filters (Lowpass Butterworth, Holoborodko), local polynomial fitting (Savitzky-Golay filter, cubic splines), and Finite Elements mapping. The proposed methodology to compare the enhancement of these methods is based on the star pattern test cases used in the DIC Challenge 2.0. Hence, this study aims to obtain a general comparison of the filtering methods as opposed to case-dependent approaches. To that end, firstly, the FOLKI-D DIC code is metrologically qualified using the DIC Challenge 2.0 methodology and test cases. Then, the filtering methods are applied to the displacements and strain maps computed with the FOLKI-D code for various DIC subset sizes and filter parameters. It is found that displacement and strain map enhancements vary significantly depending on the filter. Some methods show no improvement (or worsened tradeoff in the strain case), while others achieve similar resolution tradeoffs as DIC codes using quadratic subset shape functions. Finally, it is concluded that the filter choice significantly impacts results, with higher-order lowpass Butterworth, Savitzky-Golay, and Finite Elements mapping methods showing preference over others studied.