Robust Matching by Partial Correlation

Zhong-Dan Lan 1 Roger Mohr 1
1 MOVI - Modeling, localization, recognition and interpretation in computer vision
GRAVIR - IMAG - Graphisme, Vision et Robotique, Inria Grenoble - Rhône-Alpes, CNRS - Centre National de la Recherche Scientifique : FR71
Abstract : Visual correspondence problem (matching), %including stereo matching and motion matching, is an major issue in vision. Matching can be divided in {\it feature-based} and {\it area-based} matching. In area-based matching, correlation is a common tool for the visual correspondence problem. But most classical correlation methods fail near the disparity discontinuity, which occurs at the boundaries of objects. In this report, we propose a {\it partial correlation} technique for solving this problem. We use a robust statistics tool to find the {\it good part} to be considered for applying correlation, We have compared our method with the standard and non standard techniques. Experimental results validate the approach.
Document type :
Reports
Complete list of metadatas

Cited literature [2 references]  Display  Hide  Download

https://hal.inria.fr/inria-00074047
Contributor : Rapport de Recherche Inria <>
Submitted on : Wednesday, May 24, 2006 - 2:23:12 PM
Last modification on : Wednesday, April 11, 2018 - 1:53:18 AM
Long-term archiving on: Monday, April 5, 2010 - 12:04:01 AM

Identifiers

  • HAL Id : inria-00074047, version 1

Collections

Citation

Zhong-Dan Lan, Roger Mohr. Robust Matching by Partial Correlation. [Research Report] RR-2643, INRIA. 1995. ⟨inria-00074047⟩

Share

Metrics

Record views

322

Files downloads

373