A method of underwater image segmentation based on discrete Fractional Brownian Random Field

Under the influence of the lighting condition and some character of water media, the underwater images have low contrast, unbalance gray scales, fuzzy edge of objects and large quantity of noise which will appear with the movement of vehicle. For the mentioned factors, when traditional methods are used to dispose underwater images, the regions of objects cannot be located exactly, details of objects are lost, and shapes of objects are distorted. Considering the objects detected in underwater images are often artificial, this paper proposes a method of underwater image segmentation based on the discrete Fractional Brownian Random Field by combining the character of underwater images with the fractal theory. At first, a window is set, and the centre of window is located at the position of each pixel in the image. The average of fractal dimension in the window is calculated, and it is considered as the fractal feature of the pixel at the centre of window. At last, the threshold is determined according to the graph of fractal dimension, and the segmentation is completed. By the normalization of the average absolute intensity difference on surfaces at difference scales, the number of data items used to represent the average absolute intensity difference on surfaces at difference scales is reduced, and the segmentation efficiency is improved. Finally, the results on some typical images are presented. Comparing with the results obtained by the algorithms based on Otsu and maximum entropy, it shows that the presented method is robust, and it is efficient in underwater images segmentation.