Turbulent Flow Estimation by Wavelet Transform

Turbulent flow estimation from an image sequence is challenging due to the lack of dedicated flow measurement techniques. Existing techniques estimate flowrate with high uncertainty. In this paper, a new technique based on discrete wavelet transform (DWT) is proposed. Wavelets have the advantage of decomposing flow signals into numerous levels and remove input signal noise. The flow signals are first decomposed using DWT into multiple levels, then, the wavelet coefficients are correlated by the Fast Fourier Transform (FFT) based algorithm to determine the velocity field. This wavelet-based algorithm is named as DWT-FFT. DWT-FFT was evaluated first using synthetic signals and then applied for turbulent flow estimation. The accuracy of DWT-FFT was compared to classical algorithms including direct cross correlation (DCC) and direct implementation of FFT. DWT-FFT estimated the flow with an error of 0.7%, outperforming both DCC and FFT which estimated with an error of 7.14% and 12.2% respectively.