Accelerated Super-Resolution PIV based on Successive Abandonment Method

The super-resolution PIV based on the recursive division of interrogation window has attracted interest recently. This approach needs careful error checks because results are largely affected by the pre-step displacements obtained by large interrogation windows. Hart (2000 Experiments in Fluids Vol.29-1) proposed the error check method by multiplication between correlation distributions of overlapped neighboring interrogation windows. However, the recursive calculation with ordinary cross-correlation method requires too much computational time. In order to reduce it, Hart (2000 J. Flow Visualization Vol.3-2) used image compression techniques. In this study, we use not image compression but characteristic pixel selection to accelerate the super-resolution PIV. Our PIV is based on successive abandonment (SA) method (Kaga et al. 1993 J. Flow visualization and Image Processing Vol. 1-4). In the SA calculation with the characteristic pixel selection, 1000 candidates are narrow down to only one at over 50 % of measurement points. And also difference between correct candidate and other ones becomes clear by characteristics of selected-pixel as shown in Fig. 1, so the number of error vectors is reduced. Comparing the time per velocity vector, our super-resolution PIV is 10 times faster than the former ordinary resolution PIV. In all recursive processes, error checks are carefully done using the summation of cumulative intensity difference distribution to be suitable for SA method. Another feature of our PIV is that the velocity vectors are obtained near the boundary of images and masked regions as shown in Fig. 2. Velocity vectors near the boundary of images and masked regions are calculated carefully by extrapolation of predicted displacement, pattern matching using small interrogation window and error elimination. By these procedures, High-speed high-density PIV was achieved. -16 -12 -8 -4 0 4 8 12 16 -16 -6 4 14 0 20 40 60 80 100 120 140 D/ n