A 3D Label Stereo Matching Method Using Underwater Energy Function

Stereo vision is one of the most important ways to explore the underwater world. However, due to the irregularities of underwater scenes and significant distortion in the images taken underwater which prevents accurate matching of corresponding points by using one-dimensional disparity value between images, acquiring three-dimensional information in the complex underwater environment is very challenging. In this paper, we propose an underwater stereo matching method to estimate dense disparity maps in the underwater environment and obtain three-dimensional information about the target object. First, the 3D label underwater energy function is proposed to fit the inclined and curved planes of irregular underwater objects under the condition of underwater image distortion. Then the underwater disparity refinement method is proposed to remove the mismatch disparities caused by noise underwater. Experimental results show that the proposed method outperforms existing methods in terms of subjective and objective quality. The dataset is available at: https://github.com/lv-wj/underwater-datasets.