DOI: 10.5176/2251-1911_CMCGS14.38
Authors: Iulia Cimpan
Abstract:
(Flow of Ideas)
* Image Segmentation (what it means, why we need it)
* Watershed and Separatrix Segmentation:
- Identifying critical points in the image
- Constructing separatrices starting from saddle points
* Issues in defining critical points: different number of saddles due to different grids and different choice of nearest neighbourhood
* Aim: correct number of critical points regardless of pixel shape (rectangular, hexagonal, triangular) and regardless of the number of neighbours)
- We want to apply definitions accepted in continuous domain to a discrete domain in order to find the correct number of critical points in an image
* Mesh the image into triangulated elements
* Construct basis functions on each node (vertex) of the image
* Estimate the gradient of the image using basis functions and distributional derivative
* Calculate the zeroes of the estimated gradient
* Calculate the derivative of the estimated gradient and apply ”Hessian test”
Keywords: discrete Hessian test
Updating...