Simek, Kyle LouisView Profile. (The University of Arizona), “Branching Gaussian process models for computer vision” (2016)

Simek, Kyle LouisView Profile. (The University of Arizona), “Branching Gaussian process models for computer vision” (2016) Advisor: Jacobus Kobus Barnard

Bayesian methods provide a principled approach to some of the hardest problems in computer vision—low signal-to-noise ratios, ill-posed problems, and problems with missing data. This dissertation applies Bayesian modeling to infer multidimensional continuous manifolds (e.g., curves, surfaces) from image data using Gaussian process priors. Gaussian processes are ideal priors in this setting, providing a stochastic model over continuous functions while permitting efficient inference.

The authors begin by introducing a formal mathematical representation of branch curvilinear structures called a curve tree and the authors define a novel family of Gaussian processes over curve trees called branching Gaussian processes. The authors define two types of branching Gaussian properties and show how to extend them to branching surfaces and hypersurfaces. The authors then apply Gaussian processes in three computer vision applications. First, the authors perform 3D reconstruction of moving plants from 2D images. Using a branching Gaussian process prior, the authors recover high quality 3D trees while being robust to plant motion and camera calibration error. Second, the authors perform multi-part segmentation of plant leaves from highly occluded silhouettes using a novel Gaussian process model for stochastic shape. The proposed method obtains good segmentations despite highly ambiguous shape evidence and minimal training data. Finally, the authors estimate 2D trees from microscope images of neurons with highly ambiguous branching structure. The authors first fit a tree to a blurred version of the image where structure is less ambiguous. Then they iteratively deform and expand the tree to fit finer images, using a branching Gaussian process regularizing prior for deformation. The proposed method infers natural tree topologies despite ambiguous branching and image data containing loops. This work shows that Gaussian processes can be a powerful building block for modeling complex structure, and they perform well in computer vision problems having significant noise and ambiguity.