?url_ver=Z39.88-2004&rft_val_fmt=info%3Aofi%2Ffmt%3Akev%3Amtx%3Adc&rft.title=SEPARATING+NONLINEAR+IMAGE+MIXTURES+USING+A+PHYSICAL+MODEL+TRAINED+WITH+ICA&rft.creator=Almeida%2C+Mariana+S.+C.&rft.creator=Almeida%2C+Lu%C3%ADs+B.&rft.subject=Machine+Vision&rft.subject=Machine+Learning&rft.subject=Neural+Nets&rft.subject=Artificial+Intelligence&rft.description=This+work+addresses+the+separation+of+real-life+nonlinear%0Amixtures+of+images%2C+which+occur+when+a+paper+document+is%0Ascanned+and+the+image+from+the+back+page+shows+through.%0AA+physical+model+of+the+mixing+process%2C+based+on+the+consideration+of+the+halftoning+process+used+to+print+grayscale+images%2C+is+presented.+The+corresponding+inverse+model+is+then+used+to+perform+image+separation.+The+parameters+of+the+inverse+model+are+optimized+through+the+MISEP+technique+of+nonlinear+ICA%2C+which+uses+an+independence+criterion+based+on+minimal+mutual+information.%0AThe+quality+of+the+separated+images+is+competitive+with%0Athe+one+achieved+by+other+techniques%2C+namely+by+MISEP%0Awith+a+generic+MLP-based+separation+network+and+by+Denoising%0ASource+Separation.+The+separation+results+show+that%0AMISEP+is+an+appropriate+technique+for+training+the+parameters+and+that+the+model+fits+the+mixing+process+well%2C+although+not+perfectly.+Prospects+for+improvement+of+the+model+are+presented.&rft.date=2006&rft.type=Preprint&rft.type=NonPeerReviewed&rft.format=application%2Fpdf&rft.identifier=http%3A%2F%2Fcogprints.org%2F4914%2F1%2Fmodelo.pdf&rft.identifier=++Almeida%2C+Mariana+S.+C.+and+Almeida%2C+Lu%C3%ADs+B.++(2006)+SEPARATING+NONLINEAR+IMAGE+MIXTURES+USING+A+PHYSICAL+MODEL+TRAINED+WITH+ICA.++%5BPreprint%5D+++++&rft.relation=http%3A%2F%2Fcogprints.org%2F4914%2F