WEBINAR - BMI Progress Reports 2020 // Prof. Hess's Lab: Adélie Garin - Topological Data Analysis for Image processing

Topological data analysis (TDA) methods extract features representing the number of connected components, loops and cavities of shapes at different scales through an iterative process called a filtration. In this talk, we describe several TDA tools to improve image processing (classification, comparison and reconstruction). These methods provide a global view of images, which takes into account intrinsic geometric and topological properties of the images and complement with standard methods very well. They can be used for several purposes in image analysis: classification (of neuron shapes for example), comparison (to differentiate a healthy tissue from a stroke in the brain) or to help optimisation algorithms to reconstruct deteriorated images (for example to reconstruct 3D neuron images from scanned sliced brains) by classifying and localising the most common mistakes made by those algorithms. No prerequisites are required.