"Modeling 3-D reconstruction by image rectification of stereo images acquired by cameras of unknown and varying parameters"

Glaucoma is one of the leading causes for vision loss in the United States, and through early diagnosis and continued monitoring it is possible to delay and in some cases stop the progression of the disease. Current testing technologies such as stereo fundus photography, intraocular pressure monitoring, and visual field testing can all provide useful information regarding the progression of the disease, however, they are all subject to inter- and intra- subject variability which reduces the overall accuracy of the test and resulting diagnosis. Stereo fundus photography is by far the most popular means for detecting glaucoma as it allows physicians to not only see the cupping of the optic nerve head, but additionally, they can see the thinning of nerve fiber layers which is an indicator for the onset, or progression of glaucoma. An automated algorithm for estimating cup to disc ratios, a quantity key to tracking glaucoma progression, has already been developed. The method has already been shown to have high correlation to physician generated measures over a period of many years. The algorithm in its current state uses image warping algorithms to produce the stereo pairs, the focus of this thesis will be to replace the warping algorithms with a more direct camera geometry estimation algorithm. This will allow the images to then be rectified, or more simply, the images will be transformed into the ideal stereo camera geometry which greatly simplifies the disparity map generation and automated analysis segments of the existing algorithm. Additionally, comparisons between the existing warping method and the newly implemented geometry estimation method will be made to show that the new approach will generate similar, and in some cases more accurate results than the previous implementation.