Quantitative photomosaicking of underwater imagery

Looks at some of the justification behind photomosaicking for underwater imaging. One of the primary reasons for employing photomosaicking underwater is to obtain a better global perspective in the underwater environment. Due to physical constraints underwater, it is virtually impossible to frame large objects of interest within a single picture frame. Thus there has been considerable interest over the years in taking a series of images and composing them together into a composite photomosaic of the area of interest. The metric for judging such efforts have been largely qualitative equalizing the illumination across images to eliminate differences in intensity and paying careful attention to matching features that cross image boundaries to enhance the sense of continuity across the mosaic. Unfortunately these methods yield large overall distortions as the errors tend to accumulate in a photomosaic as successive images are added. The authors present results that highlight the need for more than a qualitative metric for determining the accuracy of a photomosaic. The authors use navigation data to analyze the rate and nature of error propagation in photomosaicking. By examining the nature of the error the authors show the potential of such a technique for generating large scale photomosaics with bounded errors.