Fast and accurate rotator for rotation-based iterative reconstruction algorithms

We present an image rotation algorithm in which the values of pixels of the rotated image are computed as weighted averages of the original grid pixels, with the weights proportional to the overlapping areas of each corresponding pixel pair. The method preserves virtually perfectly the image uniformity and single pixel pulse value in the rotated image. These properties are particularly important to accurately model projections in iterative reconstruction algorithms. The matrix of pixel redistribution coefficients can be precomputed analytically and the algorithm is much faster that e.g. the widely used Gaussian rotator.