Project:
Photometric texture (PhoTex) analysis of 3D surfaces.
The
objective of the PhoTex project is to develop a new methodology for
texture classification based on the direct use of surface gradient
and surface reflectance information. This will enable the design
of image texture analysis systems that exploit and take into
account the illumination conditions and the three-dimensional
nature of texture.
A novel approach to classification has been developed. It uses photometric
stereo to determine surface gradients. We are using this technique to extract
and separate surface relief and albedo information.
This will enable classification to be performed by comparing texture features
computed directly from surface properties rather than image intensity values.
Thus we will for instance compare surface relief against surface relief rather
than pixel patterns against image data.
Such an approach can:
provide true surface-rotation invariant (as opposed image-rotation
invariant) classification,
compensate for illumination changes that cause catastrophic
failure of conventional texture classifiers, and
improve the accuracy of texture classification and defect
detection.
Enter
PhoTex project home page for more details, click
here.
Applications
using photometric stereo, click here.