Comparison of Methods for Horizon Line Detection in Sea Images

摘要

Four algorithms designed to find a horizontal line separating sea from sky in real-life conditions were implemented and compared by their accuracy and relative speed. The algorithms selected were H-COV-LUM, based on regional covariances in luminance images, H-HC, based on edge detection and the Hough transform, H-LSC, based on maximal edge detection and the least squares method and H-MED, based on median filtering and linear regression. Real-life images were used for comparison. The most accurate line with respect to angular error was obtained by using the H-HC algorithm and with respect to the error in the position of the line by H-COV-LUM, whereas the highest speed was achieved by using the H-LSC method.