The script that that does the job of finding the frame coordinates of all (visible) surveyed reference points on a single frame was developed and tested. Figure G presents a combined output of the developed algorithm showing all major steps of it, while Figs.H-K repeat each plot shown on Fig.G in a larger scale. Staring from the original image (Fig.H and the top left plot on Fig.G) the algorithm proceeds with splitting it into three color components (RGB) and plotting the contour map of intensity of one of components. It should be noted that the reference points appearing white produce the highest intensity in each of the three components, so that any of them could be used in the further analysis. However, it turned out that red and blue components exhibit a more robust behavior.

Figure G. Analyzing the single frame of video.

The contour map for a blue component appears on the top right plot of Fig.G (Fig.I). The algorithm further finds an average intensity for the entire frame and a standard deviation (sigma). The previously computed contour map is then corrected by subtraction an average plus two sigma from each pixel. The results are shown on the bottom left plot of Fig.G (Fig.J). The further correction involves getting rid of the time stamp and finding the centroid of each group of the remaining pixels representing the surveyed reference points. The final results are shown on the bottom right plot of Fig.G (Fig.K). The (u,v) coordinates of each surveyed reference points are also shown for convenience (they are available in the digital format for the further processing as well).

As seen, for this specifically chosen sample of the video data with almost invisible (even by a naked eye) pop-up points, the straightforward algorithm does not detect them as easy as the surveyed reference points. However, if the area where the popped-up points is localized and the corresponding transformation is found (like shown in Fig.F), then the pop-up spots would be easier to detect with respect to the intensity of the surrounding pixels. In practice, for the high-quality video, even this straightforward algorithm assures finding not only the surveyed reference points, but the pop-up points as well.

Figure H. Original image. | Figure I. Intensities of the blue component. |

Figure J. Finding the peak intensity spots. | Figure K. Finding surveyed reference points. |