Please use this identifier to cite or link to this item:
https://scholarbank.nus.edu.sg/handle/10635/172328
DC Field | Value | |
---|---|---|
dc.title | COLOR IMAGE SEGMENTATION USING NEURAL NETWORKS | |
dc.contributor.author | CAO DEMEI | |
dc.date.accessioned | 2020-08-11T10:12:03Z | |
dc.date.available | 2020-08-11T10:12:03Z | |
dc.date.issued | 1996 | |
dc.identifier.citation | CAO DEMEI (1996). COLOR IMAGE SEGMENTATION USING NEURAL NETWORKS. ScholarBank@NUS Repository. | |
dc.identifier.uri | https://scholarbank.nus.edu.sg/handle/10635/172328 | |
dc.description.abstract | Segmentation plays a· significant role in image analysis. The goal of segmentation is to divide a given image into regions so that the desired objects can be extracted from the background. Color as a feature in image segmentation is increasingly used because of the information that it provides. An overview of current image segmentation techniques is first presented. Color difference measures, data clustering methods and artificial neural network models are reviewed. A survey of different approaches in color image segmentation suggests that alternative solutions are possible and necessary to improve the performance of standard color segmentation techniques. Algorithms for color clustering based on Kohonen's self-organizing map (SOM) and image segmentation are developed and implemented in this thesis. A neural network approach which is based on the SOM combined with multireso!ution processing is proposed and investigated. The basic algorithm is divided into three parts: pre-processing which transforms raw RGB data into L*u*v* uniform color space, color clustering followed by a coarse segmentation, and then post-processing leading to the final segmentation. A multiresolution stage at the front end to incorporate neighborhood information for subsequent segmentation is used to improve performance. A visual neural network system (VNNS) is constructed to deal with a class of images, only needs to train one or several images in the same category. It has been generalized to do segmentation without training, just based on a map trained by sample images from different categories. Good segmentation results are obtained for various color images. | |
dc.source | CCK BATCHLOAD 20200814 | |
dc.type | Thesis | |
dc.contributor.department | ELECTRICAL ENGINEERING | |
dc.contributor.supervisor | S.H. ONG | |
dc.description.degree | Master's | |
dc.description.degreeconferred | MASTER OF ENGINEERING | |
Appears in Collections: | Master's Theses (Restricted) |
Show simple item record
Files in This Item:
File | Description | Size | Format | Access Settings | Version | |
---|---|---|---|---|---|---|
b20223407.pdf | 8.12 MB | Adobe PDF | RESTRICTED | None | Log In |
Google ScholarTM
Check
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.