报告题目：Enhanced random walk segmentation methods and the related applications for biomedical images
报告人：边昂博士(Department of Computer Science，University of Münster)
Image segmentation is one of the fundamental problems in biomedical applications and is often mandatory for quantitative analysis in life sciences. In many cases, the use of semi-automated techniques is convenient, as those approaches allow to incorporate domain knowledge of experts into the segmentation process. The random walk framework is among the most popular semi-automated segmentation algorithms, as it can easily be applied to multi-label situations. However, this method requires a parameter and its optimal value is not easy to find. It also requires manual input on each individual image and, even worse, for each disconnected object. To overcome these drawbacks, two adaptive approaches will be presented to automatically setting this parameter respectively for Gaussian and multiplicative speckle noisy data. We also extended the random walker framework with a seed generation scheme. Thus only few manual labels are needed to generate a sufficient number of seeds for reliably segmenting multiple objects of interest, or even a series of images or videos from an experiment. The usefulness and performance of our proposals are evaluated on both synthetic as well as real-world biomedical image data. In addition, random walk is introduced into our colliding larvae tracking framework with a precious shape descriptor. This work has achieved the best results among all the state of the art techniques.
Ang Bian received her Bachelor degree in Mathematics from Sichuan University in 2010, and Master degree in Computer Science from Sichuan University, in 2013. Since then, she is a PhD student for Pattern Recognition and Image Analysis (PRIA) at the University of Münster. Her research interests include statistical model based image processing and biomedical data analysis.