A method for automatic image analysis of coke texture has been develop
ed. The method uses an optical polarizing reflecting light microscope,
with automatic sample movement. A series of images sized 405 x 270 mm
is captured using a CCD video camera. These images are subdivided int
o six squares at 135 x 135 mm each. For each of these squares an image
displaying the contours of the optical domains is made. This contour
image is based on either thresholding or a grey level gradient. The st
ructure is then characterized by the mean density and preferred direct
ion of the contours in these square images. The method thus yields two
values: A coarseness index, and a fibre index. These values do not re
quire any classification of the texture into a finite number of textur
e classes.Commercial anode cokes as well as needle and pitch coke have
been characterized. Reproducible results were obtained after 100 imag
es (600 squares) were analysed. The differences between the commercial
anode cokes are small but reproducible, while the pitch and needle co
ke deviate strongly.