You need to know the distance between those two points. I attachted part of a presentation where I did just that. I placed three "calibration blobs" at the start of the video. Masked the image for green, then red, then blue to find the "blobs". Using the euclidean distance equation you provided I took the known centroid-centroid distance divided by the pixels for each color mask and then averaged the results. Let me know if you have any further questions but you are almost there!
