The advanced noise model for an IDN based restoration of black and white pictures captured by a camera with CCD/CMOS sensor
Daniel Kekrt, Milos Klima · 2008
The deterministic blurring and noising in pictures captured by a camera with CCD/CMOS sensor can be fairly simulated as the true image transmission through some kind of ISI channel, with specific 2D impulse response (blurring) and consequently through certain random IECS-ML channel (noising). Hence for purposes of image restoration we can use the maximal a posteriori probability (MAP) criterion based iterative detection network (IDN), that is powered by the noise model and contains a number of mutually concatenated functional blocks so-called soft inversions (SISOs). This cellular structure makes an IDN suboptimal but also numerically very simple and practically applicable in contrast to an unviable optimal (single-stage) MAP detector. This paper is dedicated to the detailed description of the comprehensive noise model that controls the IECS-ML channel and that is applied in the IDN front-end, so-called the soft output demodulator (SODEM). There will be introduced three different examples of IDN front-ends for simpler IDNs, designed only for recovery of black and white images.