Vectorization, Decomposition, and PDE Identification in Imaging and Inverse Problems

Explore cutting-edge topics in imaging and inverse problems through this 58-minute virtual seminar presented by Dr. Sung Ha Kang from Georgia Tech. Delve into three key areas: image vectorization, image decomposition, and PDE identification. Learn about a mathematically founded silhouette vectorization algorithm that converts bitmap images to scalable vector files using cubic Bezier polygons and perfect circles. Discover a convex non-convex variational decomposition model for separating images into piecewise-constant, smooth homogeneous, and noisy/texture components, with applications in denoising and shadow removal. Examine methods for identifying differential equations from discrete time-dependent data, addressing challenges such as noise sensitivity and nonlinearity. Gain insights into numerical PDE techniques for stable identification of underlying PDEs. This SIAM-IS seminar offers a comprehensive overview of advanced imaging and inverse problem techniques, suitable for researchers and professionals in applied mathematics and computer vision.