Continuous Modeling Perspective for Imaging Science

Abstract:   The regularizer, which incorporates prior knowledge of images, serves as a cornerstone of inverse imaging problem modelling. In this talk, we will begin by reviewing the classical regularizers, including total variation regularizer, low-rank regularizer, and nonlocal regularizer. We then will discuss the limitations of classical hand-crafted regularizers (e.g., expressive capability, applicability, and flexibility). To address the above limitations of classical regularizers, we suggest a unified Continuous Modeling Perspective for imaging science,  which continuously represents discrete data by elegantly leveraging tiny neural networks. This paradigm allows us to readily deconstruct and reconstruct the classical regularizers. Extensive experiments demonstrate the promising performance of the continuous modeling perspective.

CV:   电子科技大学教授、博导，入选四川省学术和技术带头人、四川省天府青城计划。第一/通讯在权威期刊SIAM 系列(SISC和SIIMS)和IEEE系列(TPAMI、TIP、TSP和TNNLS等)及顶会CVPR等发表研究工作。研究成果获四川省自然科学一等奖、四川省科技进步一等奖、计算数学会青年优秀论文竞赛二等奖。主持国自然面上项目、四川省杰出青年科学基金项目、华为项目。