Variational PDE Methods for Image Processing

Course project for Variational PDE Methods for Image Processing. I implemented and compared several numerical approaches for inverse imaging problems:

• Tikhonov-regularized denoising with gradient descent, finite differences, and finite elements. Download report
• Rudin-Osher-Fatemi denoising with gradient flow, lagged diffusivity fixed point iteration, and Chambolle projection. Download report
• Split Bregman for L1-regularized denoising, inpainting, and compressive sensing. Download report

• Geodesic active contours and Chan-Vese segmentation models. GAC report

  Segmentation report

Noisy image and denoised image.

Polluted image and inpainted image.

Image segmentation with the geodesic active contour model.