Recursive Super-Resolution PIV

Accuracy and resolutions are the two main objectives in PIV. The so-called recursive or super-resolution PIV is used to achieved the objectives. It is based on beginning with a relatively larger window (e.g., 128$\times$128 or 64$\times$64 pixels) and ending with a smaller window (e.g., 32$\times$32 or 16$\times$16 pixels). Not only does the iterative process increase the spatial resolution by decreasing the window size, it also shifts the corresponding window in the second image [$f_2$ in equations (1) and (2)] based on the displacement calculated from previous iteration to obtain better correlation. The process therefore increases the spatial resolutions with the increased number of vectors and decreased size of window. A better-correlated result also means better accuracy in the correlation process. During the iterative process, spurious vectors are removed using median filter (discussed later). The Kriging method is used to interpolate the removed error vectors from the previously obtained velocity field with a larger window.

Figure 2: mpiv result from the recursive method with the 'cor' option.

Figure 2 shows the estimated velocity vectors using the correlation algorithm (with the `cor' option) with the recursive process. The MATLAB commands are listed below:

>> im1 = imread('image1.bmp');
>> im2 = imread('image2.bmp');
>> [xi, yi, iu, iv] = mpiv(im1, im2, 64, 64, 0.5, 0.5, 20, 20, 1,'cor', 4, 1);
>> [iu_ft, iv_ft, iu_ip, iv_ip] = mpiv_filter(iu, iv, 2, 2.0, 3, 1);

The window size started from 64$\times$64 pixels and ended at 8$\times$8 pixels. Similar results can also be obtained if the MQD algorithm is used (using the `mqd' option in mpiv).