Automatic detection of images containing nudity : Image detection using artificial neural networks and statistical methods

Examensarbete för masterexamen

Please use this identifier to cite or link to this item: https://hdl.handle.net/20.500.12380/76159
Download file(s):
File Description SizeFormat 
76159.pdfFulltext2.16 MBAdobe PDFView/Open
Type: Examensarbete för masterexamen
Master Thesis
Title: Automatic detection of images containing nudity : Image detection using artificial neural networks and statistical methods
Authors: Carlsson, Andreas
Eriksson, Andreas
Isik, Mikael
Abstract: This thesis discusses the possibilities of detecting images containing nudity using computer algorithms. We are only focusing on sexually explicit images. Our approach is to extract features such as skin, faces and regions, which can be used to classify images. We have investigated the advantages of the two color spaces RGB and IHLS when detecting skin. The difference in performance between the two, are illustrated in ROC graphs. The technology used is artificial neural networks, statistical methods and advanced image processing. Artificial neural networks are used for skin pixel segmentation, face detection, and image classification. Gaussian mixture models have been tested, but was too computationally heavy and was also outperformed by artificial neural networks. The separate parts performs well, but our approach using an artificial neural network with features as input does not perform as well as expected in its current state, and needs some modifications, which are proposed in the section future work. The research problem was proposed by NetClean Technologies Sweden AB.
Keywords: Information Technology;Informationsteknik
Issue Date: 2008
Publisher: Chalmers tekniska högskola / Institutionen för tillämpad informationsteknologi (Chalmers)
Chalmers University of Technology / Department of Applied Information Technology (Chalmers)
Series/Report no.: Report - IT University of Göteborg, Chalmers University of Technology and the University of Göteborg : 2008:083
URI: https://hdl.handle.net/20.500.12380/76159
Collection:Examensarbeten för masterexamen // Master Theses



Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.