Share Email Print

Proceedings Paper

Textured surface identification in noisy color images
Author(s): Mehmet Celenk
Format Member Price Non-Member Price
PDF $17.00 $21.00

Paper Abstract

Automatic identification of textured surfaces is essential in many imaging applications such as image data compression and scene recognition. In these applications, a vision system is required to detect and identify irregular textures in the noisy color images. This work proposes a method for texture field characterization based on the local textural features. We first divide a given color image into n multiplied by n local windows and extract textural features in each window independently. In this step, the size of a window should be small enough so that each window can include only two texture fields. Separation of texture areas in a local window is first carried out by the Otsu or Kullback threshold selection technique on three color components separately. The 3-D class separation is then performed using the Fisher discriminant. The result of local texture classification is combined by the K-means clustering algorithm. The texture fields detected in a window are characterized by their mean vectors and an element-to-set membership relation. We have experimented with the local feature extraction part of the method using a color image of irregular textures. Results show that the method is effective for capturing the local textural features.

Paper Details

Date Published: 26 June 1996
PDF: 7 pages
Proc. SPIE 2753, Visual Information Processing V, (26 June 1996); doi: 10.1117/12.243580
Show Author Affiliations
Mehmet Celenk, Ohio Univ. (United States)

Published in SPIE Proceedings Vol. 2753:
Visual Information Processing V
Richard D. Juday; Stephen K. Park, Editor(s)

© SPIE. Terms of Use
Back to Top
Sign in to read the full article
Create a free SPIE account to get access to
premium articles and original research
Forgot your username?