Effect of interpolation on noise propagation from images to DIC displacement maps

Abstract

International audienceThis paper introduces and validates a new prediction of sensor noise propagation from images of randomly marked surfaces to displacement maps obtained by Digital Image Correlation (DIC). Images are indeed often affected by sensor noise, which propagates to DIC output. We consider here the 2D global DIC (G-DIC), for which this output is the in-plane displacement calculated at a set of nodes. Predictive formula for the resolution of the displacement at these nodes is already available in the literature. The contribution of the present paper is to revisit this formula to take into account the interpolation required by sub-pixel displacement. A generalization is also proposed to predict the displacement resolution throughout the field of view. It is then extended to several kinds of DIC. The correlation procedure is thoroughly described in order to emphasize the role of the interpolation. A numerical assessment on synthetic data validates the new prediction and shows the improvement brought about by the proposed formula

Rendering Deformed Speckle Images with a Boolean Model

“…A full comparison of both methods in the context of the assessment of DIC algorithms is beyond the scope of this paper. Comparing the retrieved displacement field with respect to the prescribed displacement field indeed depends on various factors such as the systematic bias modeled by a Savitzky-Golay filter [43], error caused by the interpolation scheme needed by DIC itself [7,8], the shape of the speckle patterns [29,30], and quantization noise. It is thus difficult to isolate the error solely caused by the process rendering the input speckle images.…”

“…DIC assessment is, however, likely to be affected by the interpolation scheme. In order to reduce the influence of interpolation, a possibility is to generate low-resolution images by binning pixels of an interpolated high-resolution deformed speckle image [4,6,7,10,18,31,41].…”

“…It is, therefore, of prime importance to have at disposal artifact-free artificial deformed speckles to be certain that only the errors due to the DIC programs are really estimated. There is indeed an active community concerned by the assessment of the metrological performance of DIC programs, as illustrated by numerous recent papers on this problem [2,7,8,11,13,16,18,26,27,38,41,52] for instance. The Society for Experimental Mechanics (SEM) proposes such ground-truth databases within the framework of the DIC challenge [47], providing the researchers with synthetic speckle images rendered by several methods of the literature, briefly reviewed below.…”

On the Propagation of Camera Sensor Noise to Displacement Maps Obtained by DIC - an Experimental Study

“…First, an estimate of the average displacement resolution over the observed zone was proposed, e.g., [2,39]. Later on, local predictions of the resolution were introduced; see for instance [35,36,18,4]. However, their validation using experimental data has not been carefully studied.…”

NanoJ: a high-performance open-source super-resolution microscopy toolbox

“…2c). The interpolation process will, however, change the noise properties of the resulting dataset (20), which can have an influence on further analysis requiring specific assumptions in noise properties such as Maximum Likelihood Estimation (MLE) single-molecule localization analyses. For the specific case of SMLM datasets with sparse blinking, An artificially large drift was applied computationally in order to make it visible for figure rendering.…”