Segmentation of underwater multispectral images with applications in the study of coral reefs

The monitoring and studying corals reefs can be used to measure the effects of contamination in the ocean. An essential preliminary step for recognizing and analyzing corals reefs in an image is image segmentation. The main goal of image segmentation is to divide an image into parts that have a strong correlation with objects or areas of the real world contained in the image.; The segmentation algorithms used in this work can be divided in two areas: boundary detection---segmentation only across contour boundaries created by the objects and homogeneity analysis---segmentation across the quantification of the homogenous regions created by the objects. Two new algorithms based on these two areas were developed: new Hough transform for boundary detection algorithm (NHTBDA) and Hurst coefficient/Hough transform algorithm (HCHTA). The developed algorithms were compared with other two homogeneity based algorithms presented in the literature. The performance of the algorithms and the results of the segmentation are tested and analyzed in: underwater multispectral images generated in a controlled environment, underwater multispectral images taken by an Autonomous Underwater vehicle (AUV), and remotely sensed images.