Visual Saliency

Contact person: INRIA Imedia
Synopsis:

• We will investigate and develop saliency operators for images and video sequences, e.g. based on local descriptors with point-of interest detection and description. Once salient regions have been detected they have to be represented by visual features and their organization (relations) has to be described. We will work on perceptually motivated high-level features in computer vision such as symmetry, anisotropy, regularity, complexity, symmetry of textures and shapes, on multi-resolution behavior and also spatio-temporal based saliency features. We will also focus on combined representation of salient image information through region and point of interest.
  

Keywords:

• local descriptors, symmetry, regularity, content-adaptive spatio-temporal wavelets.
  

Additonal info: E-Team's webpage