Joint blind deconvolution, fusion and classification of multi-frame imagery

We propose a novel method for blind reconstruction of multi-frame images that employs multi-frame fusion and classification based constraints for recursive inverse filtering. In particular, recursive FIR filtering is applied to successively taken scan images with the same spatial resolution. The outputs of the filters are then fused and classified jointly and separately to a predetermined number of support region levels. Finally, the error between the classified output of each filter and their fusion classification is used to control the adaptation of the corresponding FIR filter parameters. Results for real MRI scan images as well as for simulated multi-frame images demonstrate the feasibility of the method and its potential for blind image reconstruction and classification.