Interpolation Problems in Image Reconstruction

The nature of interpolation required during image reconstruction using Filtered Back-Projection has been studied. It has been found that the usual method, called linear interpolation, may easily be refined to a better method, which we call trapezoidal approximation. The error associated with this method has been analyzed in the frequency domain for circularly symmetric objects. There is moderate distortion and diminshed response at high frequencies, with the amplitude envelope falling to about 60% of ideal at the Nyquist fre-quency for 100 sampling points. The error is reduced with more sampling points.