Performance of digital image velocimetry processing techniques

Digital particle image velocimetry (DPIV)-processing techniques have become increasingly more sophisticated in recent years. However, much work is still done using standard traditional methods of analysis. This paper investigates several traditionally based techniques for cross-correlation image processing in terms of computational efficiency and measurement accuracy. Direct spatial domain correlation, standard fast Fourier transform (FFT) correlation, a dynamic FFT correlation technique, and a new hybrid correlation method are discussed and evaluated. In addition, a particle-tracking velocimetry scheme based on that of Cowen and Monismith (1997) is examined in the same context as the DPIV methods. A detailed examination of the behaviors of each correlation method reveals that direct spatial domain correlation is more accurate than FFT-based methods, with the standard FFT correlation showing the weakest performance. Using the more robust methods (dynamic FFT and hybrid correlation), accuracy can be improved significantly over the standard FFT method in many cases, while still remaining computationally efficient. The particle-tracking algorithm studied was found to yield comparable accuracy to the DPIV routines and can provide much higher spatial-resolution possibilities.