Adaptive subset offset for systematic error reduction in incremental digital image correlation

• The effect of non-integer reference point location on DIC is analyzed.
• Adaptive subset offset ensures the subset pixels fall on integer positions.
• Sub-pixel interpolation in the reference image is totally avoided.
• Lower systematic error is produced compared with conventional method.
• The proposed method proves effective for incremental DIC in 2D and 3D cases.

Digital image correlation (DIC) relies on a high correlation between the intensities in the reference image and the target image. When decorrelation occurs due to large deformation or viewpoint change, incremental DIC is utilized to update the reference image and use the correspondences in this renewed image as the reference points in subsequent DIC computation. As each updated reference point is derived from previous correlation, its location is generally of sub-pixel accuracy. Conventional subset which is centered at the point results in subset points at non-integer positions. Therefore, the acquisition of the intensities of the subset demands interpolation which is proved to introduce additional systematic error. We hereby present adaptive subset offset to slightly translate the subset so that all the subset points fall on integer positions. By this means, interpolation in the updated reference image is totally avoided regardless of the non-integer locations of the reference points. The translation is determined according to the decimal of the reference point location, and the maximum are half a pixel in each direction. Such small translation has no negative effect on the compatibility of the widely used shape functions, correlation functions and the optimization algorithms. The results of the simulation and the real-world experiments show that adaptive subset offset produces lower measurement error than the conventional method in incremental DIC when applied in both 2D-DIC and 3D-DIC.