報 告 人：武婷婷 教授

報告題目：Blind Image Deconvolution: New Approaches

報告時間：2024年12月20日（周五）下午4:00

報告地點：靜遠(yuǎn)樓203學(xué)術(shù)報告廳

主辦單位：數(shù)學(xué)與統(tǒng)計學(xué)院、數(shù)學(xué)研究院、科學(xué)技術(shù)研究院 

報告人簡介：

      武婷婷教授，博士生導(dǎo)師，曾獲得江蘇高校“青藍(lán)工程”優(yōu)秀青年骨干教師、江蘇省“科技副總”、南京郵電大學(xué)1311人才計劃“鼎新學(xué)者”等稱號。目前是中國工業(yè)與應(yīng)用數(shù)學(xué)學(xué)會會員、江蘇省運(yùn)籌學(xué)會會員、中國數(shù)學(xué)會會員，擔(dān)任JMIV, JIMO等雜志及會議審稿人。主要從事優(yōu)化算法及其在信息科學(xué)、統(tǒng)計等領(lǐng)域中的應(yīng)用研究，目前發(fā)表國內(nèi)外知名期刊論文65篇，研究成果發(fā)表于《SIAM Journal on Imaging Sciences》、《Journal of Scientific Computing》以及《IEEE Transactions on Neural Networks and Learning Systems》等國際權(quán)威期刊上。

報告摘要：

      Blind image deblurring is challenging task in image restoration. In deblurring, accurate blur kernel estimation in the first stage is crucial, but errors often lead to artifacts. A robust model is proposed that decomposes the blur kernel into deterministic and random components, improving estimation through iterative refinement and utilizing the Lp-norm for better restoration. In addition, a fractional-order variational model is introduced to reduce ringing artifacts, while the Patch-wise Minimal Pixels prior distinguishes between clean and blurred image blocks. These approaches improves restoration quality and has proven convergence, demonstrating superior performance in blind image deblurring.