In the original exported nprguy ETF image, I packed three arrays into the RGB color values.

This may have removed critical information, so I am re-running the generation to export the
raw values as text.  Below is the mapping between these raw values (e->p[][].tx, .ty and .ag)
and what was originally exported.

When referring to ETF data from now n, if it is important to distinguish which set you are
using, refer to them as either "Mini ETF" or "Full ETF"...

The data files are:

nprguy-etf-tx.txt
nprguy-etf-ty.txt
nprguy-etf-mag.txt

Anyone who wants to get into this more deeply may also be interested in

nprguy-fdog-tx.txt
nprguy-fdog-ty.txt
nprguy-fdog-mag.txt

The generated drawing without any flow data is

nprguy-drawing.txt (and nprguy-drawing.jpg)


Here is the mapping I used with the original experiments:

  {
#define n(x) ((x)<0 ? 0 : ((x) > 255 ? 255 : (x)))
    int i, j;
    fprintf(stdout, "# ImageMagick pixel enumeration: %d,%d,255,rgb\n", width, height);
    for (i = 0; i < height; i++) {
      for (j = 0; j < width; j++) {
        fprintf(stdout, "%d,%d: (%d,%d,%d) #\n", j, i, n(128+(int)floor(127.0*0.97*e->p[i][j].tx/e->max_grad)),
                                                       n(128+(int)floor(127.0*0.97*e->p[i][j].ty/e->max_grad)),
                                                       n(  (int)floor(255.0*0.97*e->p[i][j].mag/e->max_grad)));
      }
    }
  }