2002 年 61 巻 2 号 p. 90-96
We devised a new three-dimensional eye movement image analysis technique using a commercialized infrared CCD camera, a personal computer and public domain software. Image analysis was performed automatically using the public domain software NIH Image. For analysis of horizontal and vertical components, the XY center of the pupil was calculated. For analysis of the torsional component, the whole iris pattern, which was rotated each 0.1 degrees, was overlaid with the same area of the next iris pattern, and the angle at which both iris patterns showed the greatest match was calculated. Rotational axis of the eyeball was calculated from three-dimensional data using the coordinates transformation technique. As a result of the analysis of artificial eye movement images, the average error of the torsional component was 0.06+/-0.09 degrees for each 1 second, and was 0.31+/-0.46 degrees for 5 seconds. Using this technique, it is possible to inexpensively perform three-dimensional eye movement analysis and rotational axis analysis from video images recorded by many types of infrared CCD cameras.