Detection of underwater objects by improved MRF

To obtain better detection results of underwater objects, this paper proposes an algorithm for underwater objects detection based on improved MRF. When the distribution of sea-bottom reverberation areas is represented by Gamma distribution model, the established parameters of three-class spatial neighborhood MRF model and level interacting model are applied to the three-class segmentation in hierarchical MRF, then the final precise three-class segmentation results are obtained. The proposed algorithm is tested by original images and artificial images with sea-bottom reverberation areas described by Gamma distribution model. The experimental results are compared with those obtained by the previous algorithm, in which sea-bottom reverberation areas are described by Weibull distribution model. It is demonstrated that the proposed detection algorithm is more accurate and faster.