Title :
Texture classification based on feature symbol random field
Author :
Ma Xiaochuan ; Fu, Xingweng ; Hou, Chaohuan
Author_Institution :
Lab. of Integrated Digital Syst., Acad. Sinica, Beijing, China
Abstract :
This paper suggests a set of texture features based on a 2-D symbol description and the calculation of co-occurrence matrices. A hierarchical algorithm based on these features is also proposed. This algorithm gives out the correct classification rate as 96.5% while the algorithm based on the gray level co-occurrence matrix features as 67.5% and that based on the bank energies as 75%
Keywords :
feature extraction; image classification; image texture; matrix algebra; 2-D symbol description; co-occurrence matrices; feature symbol random field; hierarchical algorithm; texture classification; texture features; Channel bank filters; Filter bank; Finite impulse response filter; Gabor filters; Image databases; Pixel; Production; Spatial databases; Statistics; Testing;
Conference_Titel :
Signal Processing Proceedings, 1998. ICSP '98. 1998 Fourth International Conference on
Conference_Location :
Beijing
Print_ISBN :
0-7803-4325-5
DOI :
10.1109/ICOSP.1998.770757
