|zpravodaj ČSKI - leden 2004 [ pdf ]|
datum: 20.1.2004 v 14:00
název: Image Fusion via Multichannel Blind Deconvolution
přednášející: Filip Šroubek
místo konání: UTIA
souhrn: We face a fusion problem of recovering an undegraded image from a set of degraded images. Depending on the type of the degradation, a formalized system of the most common fusion problems is built. The unknown degradation is additive noise and space-invariant blurs modeled by convolution.The fusion process is then referred to as multichannel blind deconvolution and it frequently occurs in microscopy imaging, remote sensing, astronomical imaging, etc. Fusion is handled in two distinct ways: energy minimization problem and maximum a posteriori estimator. A better restoration performance was achieved in comparison with previously proposed multichannel blind approaches. Primarily, enhanced noise robustness was observed.
The stochastic approach handles correctly situations, when the blur size is overestimated. Moreover, translation of the channel misregistration up to a certain extent can be automatically removed in the restoration process.