| Use of adaptive optics with scanning laser ophthalmoscopes (AOSLOs) has allowed for in vivo, noninvasive imaging of the human rod and cone photoreceptor mosaic. This modality could prove to be a valuable tool for clinicians in early diagnosis of retinal disease as well as provide invaluable incite for researchers. In order for these instruments to become practical in a clinical environment, many challenges must be overcome. Involuntary eye motion makes the use of AOSLOs particularly difficult as it increases imaging time, post-processing time, data storage requirements, and, most importantly, subject's chances of retinal damage due to light exposure.;The goal of this thesis is to develop a real time eye tracking and compensation system capable of overcoming slow eye drift. Data acquisition and synchronization software and electronics were developed for use with an AOSLO. A motion estimation technique based on normalized cross correlation NCC accelerated by CUDA enabled graphics cards was used as a basis for the tracking system. Motion prediction methods were developed and evaluated in order to increase the system bandwidth. Specifically, linear and quadratic extrapolation, discrete cosine transform extrapolation, and Kalman filtering techniques were used. These tracking methods were evaluated using simulated motion and real subjects. |
