MULTILEVEL CONVOLUTION FILTERING TECHNIQUE FOR DIGITAL LASER-SPECKLE-VELOCIMETRY

A new evaluation method for velocity measurements using digitized, single exposed speckle images is presented. The method is based on a convolution filtering technique used on different levels. Beginning with the computation of a small number of velocity vectors on the coarsest level, the solution is determined step by step on the finer levels, and the number of points is squared from one level to the next. On the coarsest level the vectors are computed with high accuracy, and good approximation is obtained through interpolation of the solution on the next, finer level. Preprocessing of the images considerably improves the accuracy and evaluation speed of the measurement. The computation of the displacement vectors on the finest level without interpolation shows that the number of erroneous vectors computed during the binarization of the images can be reduced by up to 70% Using the convolution filtering technique on three levels allows for a further reduction of erroneous vectors by up to 40% Use of smaller kernels and reduction of the kernel and the image area after every interpolation step reduces the computation time for a velocity vector field to 50% compared to the one-level algorithm.