Advancing image modelling using level-sets : simulation, convergence, practical inference analysis

The supporting material includes figures and tables detailing the use of level-sets within image modelling in the thesis titled 'Enhancing spatial image analysis: modelling perspectives on the usefulness of level-sets'. Specifically the use of level-sets for noise removal in images, enhancing the existing Adaptive Median Smoother technique. Level-sets are then used to represent images as graphical models. These graphical model representation are then used to train model image of the fibrin networks of healthy and asthmatic patients to allow for inference to be drawn. Lastly, it is shown how level-sets can be used to improve a technique called D-RISE for improved explainability of deep learning models.

本辅助材料包含图表，详细阐述了题为《增强空间图像分析：水平集（level-sets）应用价值的建模视角》的论文中，水平集在图像建模中的具体应用。具体而言，研究首先将水平集用于图像降噪任务，对现有的自适应中值平滑器（Adaptive Median Smoother）技术进行优化升级。随后，研究人员借助水平集将图像表征为图模型（graphical models）。基于此类图模型表征，可针对健康受试者与哮喘患者的纤维蛋白网络训练图像模型，进而开展推理分析以获取相关结论。最后，本文还展示了如何利用水平集改进D-RISE技术，以提升深度学习模型的可解释性。