Vries, de, Niek (2019) User-driven image co-segmentation. Bachelor's Thesis, Computing Science.
|
Text
bCS_2019_deVriesN.pdf Download (19MB) | Preview |
|
![[img]](https://fse.studenttheses.ub.rug.nl/style/images/fileicons/text.png) |
Text
Toestemming.pdf Restricted to Registered users only Download (116kB) |
Abstract
In this bachelor thesis, an application has been implemented that can be used to interactively perform image co-segmentation. The main steps in this co-segmentation pipeline consist of a superpixel extraction step using SLIC, a feature extraction step including features based on color values, SIFT, and HOG, computing feature vectors for these superpixels, and using these feature vectors in two different co-segmentation approaches consisting of a k-means clustering method and a graph-cut method. For segmentation using graph-cut, the results are evaluated in the form of uncertainty scores for every superpixel. The pipeline was implemented in a way that each individual step can be performed separately. These steps are integrated in a graphical user interface, providing a user with tools to conveniently select the input images, set the parameters, provide input markings on the images, and view the results of the intermediate steps. Methods to compare the segmentation results to a ground-truth image are also provided. The application was tested on the CMU-Cornell iCoseg Dataset and showed that it was able to produce good segmentations especially on images where there is a high contrast between the subject to be co-segmented and the background.
| Item Type: | Thesis (Bachelor's Thesis) |
|---|---|
| Supervisor name: | Kosinka, J. |
| Degree programme: | Computing Science |
| Thesis type: | Bachelor's Thesis |
| Language: | English |
| Date Deposited: | 11 Jul 2019 |
| Last Modified: | 12 Jul 2019 11:14 |
| URI: | https://fse.studenttheses.ub.rug.nl/id/eprint/20100 |
Actions (login required)
 |
View Item |