Abstract

We have developed a robust method for automatic image segmentation based on a local multiscale texture description. First, we apply a bank of Gabor filters to the input, producing a joint representation of the image in the spatial and spatial-frequency domains. Then we obtain local texture descriptors, by basically computing the modulus of the filters' complex output. These texture descriptors constitute the input of a standard clustering segmentation algorithm. In this paper we present results of automatic segmentation and area computation of transmission electron microscopy (TEM) micrographs from Sb materials, containing areas featuring different crystallization. The results are highly satisfactory with a mean error of about 4% in area estimation. Since this method does not include any ad hoc feature for this particular application, an equivalent performance could be expected from other similar applications with textured micrograph images.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call