Localization of underwater objects is essential for exploring the vast ocean resources. We present a monocular-vision-based method for estimating three-dimensional (3D) position-pose of underwater circular features. Using a single camera image, underwater circular features are extracted. After that, the image distortion caused by the underwater environment is compensated, and both the orientation and position of the underwater circular features are estimated. Laboratory experiments demonstrate that the method is capable of estimating the 3D position-pose of a cylindrical object in clean water with the accuracy of orientation and position estimation of 1.33% and 4.37%, respectively. The stability of the method is also analyzed by moving and relocating the cylinder.