Abstract

Data representing the image of an original picture having high resolution are first acquired. The data are then processed in accordance with a blurring function, which can be a randomly distributed point spread function. Processing can be carried out either in object space or in spatial-frequency space. In object space a multi-dimensional convolution is required. In the spatial frequency domain, processing is performed by a pointwise product of data representing the Fourier transform of the original image and the Fourier transform of the blurring function. The resulting blurred image is thereby coded such that the original scene is no longer discernible, but can be recovered provided the blurring function is known. Additionally, this blurred image can be shown to contain the detail of the original but can be stored or transmitted as data which are compressed by compression ratios which can range from 2 to 100. The compressed data can be restored to provide an image with detail like the original by inverse filtering using data representing the Fourier transform of the blurring function.