The orientation of features in digitized scanning electron micrographs may be estimated using a technique which examines the changes in intensity in two orthogonal directions about each pixel. Typically, over 250, 000 estimates of orientation may be made generating an output image where the pixel value relates to the orientation direction of the feature at the corresponding pixel in the original image. At the same time, a rosette histogram may be generated to give a visual impression of any preferred alignment. These data may be reduced to two parameters specifying the direction and degree of orientation.
A new algorithm involving the passage of a large radius modal filter over the orientation-coded, image examines for clustering of features into domains of similar alignment. Results from the use of different filter radii will be discussed. The resulting domain-segmented image may then be processed using traditional particle feature image analysis routines to determine the size and shape etc. of the domains. Finally, the domain-segmented image may be combined with the original to produce a domain-mapped image, highlighting the various domains in different colour.
The steps outlined above may be run sequentially on each image, and batch processing of a large number of images is possible providing an automatic and objective method of measuring microfabric within images.
Tovey, N. K.; Smart, P.; Hounslow, M. W.; and Desty, J. P.
"Automatic Orientation Analysis of Microfabric,"
Scanning Microscopy: Vol. 1992
, Article 29.
Available at: https://digitalcommons.usu.edu/microscopy/vol1992/iss6/29