主题：Learning-based Adaptive Representations for Image Restoration

时间：1月7日上午9:30—10:30

地点：玉泉校区行政楼208会议室

报告人：阎若梅博士

报告摘要（Abstract）:

The purpose of image restoration is to remove the artifacts which degrade an image. They might include many forms such as motion blur, noise, artifacts from the codec etc. Adaptive image representation is of high importance for image restoration despite the fact that it has been considerably researched in the past. Numerous types of representations, e.g. wavelets, have been proposed in the past for their sparse properties. However, the basis functions in most of those methods are predefined and fixed, which will result in artifacts for various input images. My research is to provide novel learning-based adaptive representations for many modalities such as natural images. In this talk, two pieces of my work will be presented. The first one is an improvement on sparse representation for image denoising at the challenging high noise levels of the Gaussian noise. The second one is our proposed deep learning system for image blur segmentation, which is critical for extracting useful high-level regional information.

报告人简介（Short biography）:

Ruomei Yan received her B. Eng., M. Eng. Degrees in Telecommunications Engineering from Xidian University, Xi’an, China in 2007 and 2010 respectively. She is now a PhD candidate in the department of Electronic and Electrical Engineering, the University of Sheffield, Sheffield, UK. Her research interests include Image Analysis, Machine Learning for Image Processing, and Image Compression.