Comparing image processing techniques for bubble separation and counting in underwater leaks

This work aims to present a study realized to improve the counting of bubbles in underwater videos for applications of oil spill or natural exudation detection. It explains all the accomplished results through used techniques. The numbers of bubbles and their positions are compared versus human observations and with previous techniques. It shows an improvement in the detection, separation and counting of very close bubbles that are presented in terms of accuracy, sensitivity, specificity, predictivity, efficiency, Youden index and F-Score. The used video and frames are available for future comparison.