New look at maximum entropy image reconstruction

This paper presents new insights into the maximum entropy (ME) method of image restoration. It is shown that when a specific image prior probability pdf model is chosen for Bayesian MAP restoration, the resulting solution is identical to the maximum entropy result. This relationship provides a new means of evaluating the theoretical foundations of maximum entropy, and may assist in determining what class of images are best suited for ME processing. Also, a new non-iterative, closed-form approximation to the ME solution is developed. This result can reduce computational demands compared to conventional iterative algorithms. An example of the closed form restoration is presented.