Lower bound on average mean-square error for image restoration

摘要

An average mean-square error bound that is applicable to general image observation models involving degradations of blur, signal-dependent and signal-independent noise, and sensor nonlinearity is derived. A Cramer-Rao lower bound on average mean-square errors for any unbiased image restoration scheme is derived. This bound is analytically expressed as a function of degradation parameters of imaging systems. Potential performance improvements by incorporating signal-dependent noise or sensor nonlinearity into algorithmic design are discussed.