In order to improve the dynamic range, spatial resolution and measurement accuracy, a number of iterative methods have been proposed. In these methods, the three-point Gaussian peak fit of correlation value is usually used for a sub-pixel analysis. However, it is reported that the root-mean-square (RMS) error is no less than 0.1 pixels because of simple courve fitting concept, and that it requires high particle concentration or the large correlation window.
The authors have proposed the high-resolution PIV technique based on the crosscorrelation and the gradient method. In this paper, the sub-pixel displacement in the proposed technique is obtained using the spatio-temporal derivative method instead of the three-point Gaussian peak fit. The error of the proposed technique is analytically evaluated as the function of sub-pixel displacement, the particle image density and the out-of-plane displacement using Monte Carlo simulations. It is confirmed that the present method shows the higher performance both on the accuracy and the special resolution than the Gaussian fit method, and its RMS error is in the order of 1/100 pixels.