Automatic algorithm for geometric correction in subtraction radiography

Digital subtraction is currently used in radiographic studies of periapical lesions or other dental disorders that have been treated and whose progress must be evaluated over time. This paper presents an accurate and reliable automatic method for correcting the geometric distortions that are frequently present in this kind of studies. The algorithm corrects these geometric distortions comparing two images, transforming one of them (the floating image) into the other (the reference image) so that the pixels representing the same physical structure can be superimposed. Firstly, the algorithm performs an affine transformation on the floating image. Then, the resulting image is compared with the reference using a similarity measure, the correlation ratio. Finally, a standard minimization algorithm, the downhill simplex method in multidimensions, chooses transformation parameters to maximize similarity.