The accuracy and reproducibility of a global method to correct for geometric image distortion in the x-ray imaging chain

A method to correct for geometric image distortion in the x-ray imaging chain, so-called dewarping, has been developed. A global two-dimensional polynomial model of which the degree is optimized is used. The performance of the method has been tested in a number of experiments using images of a plate with a I cm spaced wire grid put against the input screen of the x-ray image intensifier (14/17/27 cm). Both offline cine film and online video images were analyzed. The accuracy of the dewarp method was derived from the acquired images and from computer-simulated distorted images. The robustness and reproducibility of the dewarp method was evaluated by means of imaging the grid in various random orientations. Three parameters describing the behavior of the algorithm were considered. One is the reproducibility of the location of a dewarped position. The second parameter is the reproducibility of the distance between two adjacent dewarped positions as a measure of the reproducibility of the size of an object under investigation. The third parameter is the reproducibility of the pixel size in the plane of the calibration plate. The major results are: the reproducibility of the location of a dewarped position was 0.01-0.04 mm for cine film and 0.04-0.07 mm for video images. The coefficient of variation of the distance between two dewarped positions was 0.04%-0.11% for cine film and 0.15%-0.18% for video images. The dewarp algorithm turned out to be fast and accurate and the distortion was removed over the whole image field down to a low random residual level. It was found that a random orientation of the grid did not affect the assessment of the distortion nor its correction. The dewarp method proved to be intrinsically robust and highly reliable. Time instability of the imaging chain was the main source of variability in the dewarp results. (C) 1997 American Association of Physicists in Medicine.