Video recognition of breaking waves

An algorithm is presented to automate the identification of breaking waves in images collected with a camera on a drifting buoy. Each image is given a score from four separate analysis techniques: brightness detection, pixel histogram, entropy (texture) analysis, and glare identification. By combining these in a composite score, potential breaking wave images are detected and the number of images requiring manual review is a small fraction of the original set. Most of the images with false breaking wave signals due to sun glare are identified and removed. The final output is the wave-breaking rate over the length of the video capture.