Multichannel Image Decoding Using a Composite Filter.

摘要

One approach for decoding an image coded by a linear-space invariant filter involves using an inverse filter. The Fourier transform of the coded image is multiplied by a function that is proportional to the inverse of the coding transfer function. With an infinite dynamic range, it is theoretically possible to recover all of the Fourier transform of the image except at the zeros of the coding transfer function. However, in practice, noise prevents recovery of the transform in regions where the transfer function is at or near zero. The problems of noise can be partially resolved by using a linear inverse filter that minimizes the mean-square error between the image and decoded image. Such filters have been implemented both optically and digitally. Still, it is not possible to recover any information below the noise level. In this Letter a technique is suggested that can, in certain cases, retrieve missing information around zeros. The additional information is obtained by coding the same image more than once with different codes. Each version of the coded image provides a dfferent information channel for recovery of the image.