Measured displacement fields constitute a rich database for the identication of mechanical behavior. A variety of methods are available today to address this problem. A formalism is proposed showing that they all revert to the minimization of a quadratic norm (or semi-norm) between measured and computed displacement elds. However, they differ in the chosen metric. Transporting image noise through Digital Image Correlation and identification procedures, the uncertainty in the measured displacement fields (i.e., their full covariance) and the resulting uncertainty in the estimated constitutive parameters is assessed. Consequently, an optimal choice of metric (with respect to image noise) is proposed so that the resulting identication uncertainty is minimized.