Clipped noisy images: Heteroskedastic modeling and practical denoising

摘要: We study the denoising of signals from clipped noisy observations, such as digital images an under- or over-exposed scene. From a precise mathematical formulation and analysis problem, we derive set homomorphic transformations that enable use existing algorithms for non-clipped data (including arbitrary filters additive independent identically distributed, i.i.d., Gaussian noise). Our results have general applicability can be ''plugged'' into current filtering implementations, to more accurate better processing data. Experiments with synthetic real raw charge-coupled device (CCD) sensor show feasibility accuracy approach.