VTM--an image-processing system for measuring ocular torsion

This paper reports a new, fast, accurate realization of an image-processing method of measuring ocular torsion (rotation of the eyeball around the visual axis) called Video Torsion Measurement (VTM). The method is to cross-correlate the two grey-level distributions of an arc of the iris from two separate images using a fast image processor card interfaced to an IBM-AT compatible computer. The card (Matrox MVP-AT) is supplied with a library of low-level functions for controlling the hardware operations of the board and the VTM system software, which is written in the C programming language, incorporates these low-level functions to interface with the MVP-AT board as well as carrying out the data-acquisition and processing algorithms. These programs: acquire an image of an iris illuminated by a single infrared (IR) light source; threshold this image in order to identify the pupil; calculate the pupil area and locate the centre of the pupil using a centre-of-gravity algorithm; record the grey-level distribution along an arc 256 pixels long at a selected radius from the pupil centre; carry out an FFT on this (interpolated) grey level distribution; store the parameters of this reference FFT and cross-correlate the comparable iral grey-level distribution from other test images of the same eye in order to determine the amount of torsional rotation of the test images relative to the reference image. This system is interactive and is designed for operation in a clinical testing situation with a minimum of operator intervention. The VTM system has a resolution of the order of 0.1 deg depending on the arc radius used and it has been validated in two ways: by using it to measure known torsional rotations of an artificial iris-like pattern and also by direct simultaneous comparison of measures on the same human iris images from VTM and those from the standard 35 mm photographic procedure of measuring torsion.