H-Index
75
Scimago Lab
powered by Scopus
JCR
Clarivate
Analytics
18%
Acceptance
Rate
call: +1.631.470.9640
Mon-Fri 10 am - 2 pm EST

Logo



eISSN: 1643-3750

Segmentation of biomedical textured images using neural networks

Michał Strzelecki

Med Sci Monit 1996; 2(4): MT505-510

ID: 500003

Published: 1996-07-01


The problem of applying neural networks models to the important problem of textured image segmentation is presented. The term "visual texture" is defined and conventional methods of texture classification are outlined. The new classification method which uses the statistical technique of coocurrence matrix in conjunction with multilayer perceptron network is developed. The proposed method is tested on sample biomedical images for which the method performs better than the classical method of multidimensional variance analysis. Also, the developed neural classifier exhibits additional advantages over conventional methods: requires less computation time, is easier to implement, and its structure is well suited for hardware realisation.

Keywords: image segmentation, texture, coocurrence matrix, neural networks



Back