Efficient subpixel image registration algorithms

摘要

Three new algorithms for 2D translation image registration to within a small fraction of a pixel that usennonlinear optimization and matrix-multiply discrete Fourier transforms are compared. These algorithmsncan achieve registration with an accuracy equivalent to that of the conventional fast Fourier transform up-nsampling approach in a small fraction of the computation time and with greatly reduced memory require-nments. Their accuracy and computation time are compared for the purpose of evaluating a translation-ninvariant error metric. © 2008 Optical Society of America