Digital image processing of underwater images

We describe the motivation and some preliminary results of a long term project to monitor a coral reef using computer vision. We show the calculation of a background image of coral without fish by median filtering of an image sequence. We illustrate the extraction of fish from images and show that a texture can pair examples of fish species from different images. We show that man-made objects in the image can be distinguished from natural objects. We propose that a control strategy to adjust image processing parameters, and sequence multiple applications of tools, is especially important for digital processing of underwater images.