Maximum Entropy Image Restoration Revisited

This paper presents a new non-iterative, closed-form approximation to the maximum entropy (M.E.) image restoration method. A fast frequency domain implementation of this closed form approach is developed for the case of circular convolutional blur. This result dramatically reduces computational demands compared to conventional iterative M.E. algorithms such as MART. Some limitations and advantages of M.E. restoration are investigated, including its dismal performance for high resolution restoration of decimated or randomly sampled blurred observations.

• Type: Conference
• Year: 2000
• Where: ICIP
• Authors: Matthew Willis, Brian D. Jeffs, David G. Long
• Keywords: Circular Convolutional Blur, Conventional Iterative M.e, ICIP 2000, Image Processing, Image Restoration Method