Discrete orthogonal polynomial restoration of images degraded by spatially varying point spread functions

This investigation evaluated the application of discrete orthogonal polynomial deconvolution (DOPD) to the restoration of two dimensional images degraded by separable spatially-varying point spread functions. The one-dimensional DOPD techniques developed for time-varying systems easily extend to the separable, spatially-varying two-dimensional case. The quality of the restoration depends on the basis set selection and on the characteristics of noise in the degraded image. This method of restoration is noniterative, and does not require matrix inversion or segmentation of the image into locally invariant PSF regions. The primary shortcoming is that the quality of the restored image depends on the adequacy of representation of the original image by a limited basis set. Initial experimental results with severely degraded noisy images are promising. Further investigation will establish the scope of practical image restoration applications