管理员登录 / English
Multi-parameter Image Approximation based on Framelet and Total Variation Regularization
发布时间: 2018-01-12     17:08   【返回上一页】 发布人:樊启斌


 

北京师范大学数学科学学院

 

周五公众报告

 

 

报告题目:Multi-parameter Image Approximation based on Framelet and Total Variation Regularization

报告人:樊启斌  (武汉大学)

时间地点:11916:00-17:00 后主楼1124

邀请人:杨大春

报告摘要:Image restoration based on variational method is one of the hot topics in digital image processing. The best-known ROF model preserves shape edges but causes stair case in the flatten part. For the remedy of this defect, Several works proposing high order regularization method have been done in the literature, but which suffered numerical inefficiency due to high nonlinearity and non-smoothness of the Euler-Lagrange equation. Considering this we propose a new multi-parameter regularization model based on framlet and total variation to restore image from blurry and noisy or corrupted observations, which utilizes edge preserving property and the adaptivity of the framlet transform to the underlying regularity of image. We discussed the solution of the proposed model by alternative direction minimization algorithm which equivalent to split Bregman iteration with an efficient and automatical multi-regularization parameter choosing strategy based on balancing principle which overcome the difficulty of toning the regularization parameter manually in numerical simulation. Complexity of the algorithm was analyzed and convergence of the algorithm was proved and the performance of this new model was shown by several benchmark numerical results.

 

 

周五学院公众报告主旨:讲解现代数学中的基本概念、重要结果及其数学思想的起源、方法的创新等,扩展我们的教师和研究生的视野、提高数学修养以及增强学院的学术氛围。邀请各方向的专家用通俗易懂的报告内容、自由互动的讲解方式展现现代数学中重要的基本概念和深刻原理,让听众可以领略到现代数学理论其实是来源于一些大学或研究生阶段就熟知的古典数学中的想法和结论。希望这些报告有助于高年级的本科生、研究生和教师更全面的认识现代数学、享受数学之美。欢迎大家踊跃参与!