An improved method for directional image smoothing based on structure tensors and vector field visualisation techniques.

Abstract : We propose an alternative diffusion technique, starting from a tensor-based method that can perform both isotropic and anisotropic smoothing using as directional information the eigenvectors of the image structures. The method employs two tensors, a structure and a diffusion tensor respectively, the novelty consisting in the manner in which we design the diffusion tensor. The developed method allows a fast implementation and a reduced number of time iterations, being discretized by a trace-based method. Testing is performed both on synthetic and real images, with quality measurements and remarks.
Document type :
Journal articles
Complete list of metadatas

https://hal.archives-ouvertes.fr/hal-00333232
Contributor : Olivier Lavialle <>
Submitted on : Wednesday, October 22, 2008 - 5:20:07 PM
Last modification on : Thursday, January 11, 2018 - 6:21:06 AM

Identifiers

  • HAL Id : hal-00333232, version 1

Citation

Cosmin Ludusan, Olivier Lavialle, Sorin Pop, Romulus Terebes, Monica Borda. An improved method for directional image smoothing based on structure tensors and vector field visualisation techniques.. Transactions on Electronics and Communications, 2008, 53 (67), pp.124-128. ⟨hal-00333232⟩

Share

Metrics

Record views

206