The presence of highlight can lead to erroneous results in several Computer Vision applications, as such as histogram equalization, objects detection, among others. Many algorithms have been developed to detect and remove highlight. In this project, I propose a very simple and effective statistical approach for highlight detection, which is based on the Mahalanobis Distance. Performance evaluation is performed in 65 images with different contexts. The experimental results so obtained indicate that the proposed technique is an effective tool for highlight detection and may lead to new alternatives for exploiting the potential of statistical analysis in digital image processing.
Matlab R2014
-
/code/algorithms
-
/code/aux_codes
-
/code/methods
evaluation_algorithms.m
Danilo Pena