Feature space dimensionality reduction by using local represent ability

【Author】

Josef Bign;Caspar Horne;

【Abstract】

<正>This paper decribes an approach to the dimensionality reduction of a feature space in an unsupervised manner. Reducing the dimensionality of the feature space is a problem which often appears in segmentation of the images.A subset of a feature set often performs better in a segmentation algorithm than including the total feature set.In this paper we propose to reduce the dimensionality of the features by using the principles of Karhunen-Loeve transform.The proposed method first evaluates the local scatter matrix for all neighborhoods of a prescribed size followed by computing the eigen vector corresponding to the largest eigenvalue.The obtained vectors are the best coordinate axes proposed by the local images.These are then processed further yielding an LMS optimal complete orthonormal(ON) coordinate system.Experimental results based on real images with a few as well as many regions are discussed.

【Keywords】

Feature space dimensionality reduction by using local representability;

References

To explore the background and basis of the node document

Springer Journals Database

Total: 8 articles

  • [1] Duda RO, Hart PE, Pattern Classification and Scene Analysis,
  • [2] Athanasids Papoulis, Signal analysis,
  • [3] Wouk,A, A Course of Applied Functional Analysis,
  • [4] Burt P J, Fast filter transforms for image processing, Computer Graphics and Image Processing, Computer Graphics (Computer Graphics)

More>>

Similar documents

Documents that have the similar content to the node document

Springer Journals Database

Total: 71 articles

More>>