Search

The automatic or semi-automatic research of archaeological findings includes several methodologies and algorithms of computer vision. The reconstruction of a scene is one of the key steps to meeting that challenge. This paper addresses a methodology for the reconstruction of underwater scenes with mosaicing techniques. The reconstruction of various scenes will be a video mosaic of sea-bottom landscapes starting from single video frames. The methodology is based on the evaluation of optic flow between frames, and its motion estimation has been evaluated on features extracted from the common areas of pairs of consecutive frames. This approach takes the motion model from a geometric projection framework. The estimation of camera movement is a second key point in the mosaicing problem. The methodology used should be robust enough to produce a good performance because of the high level of noise and turbulence involved in sea-bottom video acquisition. For this purpose, geometrical transformations have been used to map each frame into a unique big common reference frame with dimensions similar to that of the union of frames. [ABSTRACT FROM AUTHOR]