Computing oriented texture fields

  • A. Ravishankar Rao
Part of the Springer Series in Perception Engineering book series (SSPERCEPTION)


This chapter deals with visual textures that are comprised of flow-like patterns such as wood grain. These textures are characterized by local selectivity of orientation, which can vary arbitrarily over the entire image. In other words, the texture is anisotropic. Every point in the image is associated with a dominant local orientation, and a local measure of the coherence or degree of anisotropy of the flow pattern. One way of visualizing oriented textures is to think about the image intensity surface as being comprised of ridges, whose direction and height can vary continuously. We define the orientation field of a texture image to be comprised of two images, called the angle image and coherence image. The angle image captures the dominant local orientation at each point in the texture in terms of an angle, and the coherence image represents the degree of anisotropy at each point in the texture. The term orientation field is used because it is closely related to a velocity flow field, where at each point in space, a fluid element can have some velocity, which is composed of a magnitude and direction. However, the analogy is not complete, as we shall show later. The major contributions of this chapter are to present optimum methods for the computation of the angle and coherence images, and to present the results of several experiments that illustrate the importance of these images in different contexts.


Local Orientation Gaussian Filter Gradient Vector Filter Size Orange Peel 
These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.


Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.

Copyright information

© Springer-Verlag New York, Inc. 1990

Authors and Affiliations

  • A. Ravishankar Rao
    • 1
  1. 1.IBM T.J. Watson Research CenterYorktown HeightsUSA

Personalised recommendations