Restoration of Degraded Images with Maximum Entropy

Images recorded in electronic or photographic media are often degraded due to system imperfections such as diffraction effects, optical system aberrations, camera and/or object motions, defocussing, or atmospheric turbulences. In addition to these blurring effects, the recorded image is usually corrupted by various random noise effects such as the randomness of the film grain, photoelectric effects, but also random measurement or transmission errors and a further degradation caused by digitization. The general goal of an image restoration is then to model these degradations in an appropriate way which, given the recorded data, allows for estimating the original signal. Our recording model described below comprises both spatial degradation caused by blur, and random pixel degradations responsible for the recording noise. We assume a linear system transmission of the form