Computing Negentropy Based Signatures for Texture Recognition
![Thumbnail](/bitstream/handle/20.500.14043/24960/ugal_f3_2007_nr8_Lungu.pdf.jpg?sequence=4&isAllowed=y)
View/ Open
Dată
2007Autor
Lungu, Ana-Elena
Colţuc, Daniela
Frângu, Laurenţiu
Abstract
The proposed method aims to provide a new tool for texture recognition. For this purpose, a set of texture samples are decomposed by using the FastICA algorithm and characterized by a negentropy based signature. In order to do recognition, the texture signatures are compared by means of Minkowski distance. The recognition rates, computed for a set of 320 texture samples, show a medium recognition accuracy and the method may be further improved.