Windsurf (Wavelet-based INDexing of ImageS Using Region Fragmentation)



Software library

Windsurf system

Related papers



The software library includes, as one of the possible instantiations of the framework, the Windsurf system, as described in the original Windsurf paper (IWOSS'99). In summary, from the point of view of feature extraction, the Windsurf system is characterized by:

  • image representation based on the Discrete Wavelet Transform (DWT) filter in the HSV color space;
  • region segmentation based on the K-means algorithm, clustering wavelet coefficients according to the Mahalanobis distance;
  • region features include the centroid and the covariance matrix of the wavelet coefficients;
  • region comparison using the Bhattacharyya distance function on the regions representations.

Although, in its original implementation, the Windsurf system had its specific query processing models and techniques, by instantiating it within the framework we extended Windsurf so as to use all the query processing models and techniques included in the Windsurf framework.