The Swedish Navy uses remotely operated vehicles (ROV:s) for mine clearance. The visibility of underwater image sequences in for example the Baltic Sea is, however, severely limited by particles occluding the camera view, which reduces the mine detection and identification range. Spatiotemporal analysis has the potential to significantly improve the image quality and increase the detection range. This thesis describes a framework for image enhancement, object detection and object shape estimation. This framework is intended to support the identification, performed by the ROV operator, and reduce the risk of colliding. The thesis focuses on alignment of sequences where an object is severely occluded by free-floating particles. The analysis is based on the hypothesis that the image motion of particles is inconsistent with the motion of a stationary rigid body. Since the motion of a stationary rigid body can be predicted, using platform navigation data and plane parallax decomposition, a stationary rigid body can be identified in spite of small spatial support. Theories of how to progress with object shape estimation and image enhancement are presented as well as how a sensor upgrade of the ROV system could improve the analysis performance.