A Moving Window Shannon Reconstruction Algorithm for Image Interpolation

Abstract The ubiquitous problem of interpolating digital image arrays can be reliably met by a moving window Shannon reconstruction. After a discussion of the application of Shannon's theorem within a finite sampling interval, the article discusses the moving window algorithm and describes how truncation effects are avoided. The choice of the window width in which each interpolated value is computed depends upon the type of image to be reconstructed and upon the computational resources. To achieve maximum efficiency, the implementation of the algorithm can make use of look-up tables of transcendental coefficients; jitter errors introduced by the use of such tables are briefly discussed. The paper presents a comparative series of results obtained with various interpolating algorithms, both on test functions and on real images. The behavior of the moving window reconstruction is outstanding in all cases, provided that a good interpolating kernel and a proper window are adopted. A comparison is also made with the results obtained by a Shannon reconstruction based upon the entire sampling interval.