16 Jul Universal autonomous graph-based image segmentation with near-linear average complexity
This project is a generic approach to image segmentation based on the perception of images as graphs. Image segmentation is all about discovering the essential parts of an image – areas of similar characteristics. Images consist of pixels and in this project their relations form a system of differential equations, which provides information to merge pixels into larger groups of pixels. Several iterations are performed to achieve a final segmentation of the image. This technique of merging groups of pixels makes it possible to achieve a fast algorithm with near-linear average complexity. Moreover, the technique is very general, and therefore gives the ability to universally segment images with any number of segments, and do it autonomously – meaning without user input.
Category: COMPUTING — Country: DENMARK — Year: 2020
William Bille Meyling
Video:
Poster:
