Improving Metrological Performance Estimation of Digital Volume Correlation: Application to X-Ray Computed Tomography

Abstract

Background

This study reports on the performance estimation of Digital Volume Correlation (DVC) for tomographic applications. The performance of DVC can be evaluated in terms of two distinct errors: the random error, directly linked to image quality, and the interpolation error, which is the one of the most significant systematic error generated by DVC algorithms. However, the existing methods provide only a limited estimate of the interpolation error, or allow only the random error to be assessed.

Objective

A new method is proposed to evaluate the interpolation error coupled with the random error in a simple and fast way to assess the overall performance of DVC for any tomographic application.

Methods

This new method proposes to apply a rotation to the sample (instead of the usual translation) to evaluate the interpolation error. This rotational movement generates linearly varying displacement fields, and each point of a displacement field describes a distinct non-integer voxel position. As this rotation is a rigid body motion, the random error associated with tomographic noise is also taken into account.

Results

This new method can generate several thousand interpolation error measurement points in only two acquisitions, allowing a very detailed and local assessment of this error. Additionally, and compared to existing methods in the literature (repeat scan), this method does not underestimate the random error, essential for assessing the overall performance of the DVC.

Conclusions

The proposed method efficiently evaluates DVC performance by accurately assessing both interpolation and random errors through rotational sample movement, improving the reliability in DVC measurements.