Color is believed to be an essential clue in describing the world around us. The Marie Curie fellowship research focused on incorporating color information into object recognition, which traditionally only used luminance information. The success of the fellowship is illustrated by the publication of the research in the major conferences of computer vision and its excellent results on several benchmarks. The current project is a follow-up project on the Marie Curie results, and its aim is twofold. Firstly, a disadvantage of current approaches is that the applied color description is independent of the task and class of objects. Although color description will most probably improve recognition of color invariable categories, such as flamingos, it is doubtful if color helps to recognize color variable categories, such as cars. In this project we propose to combine the theory of discriminative power and photometric invariance into a singly theory. An analysis of the discriminative power of color for a particular task, should allow for a task specific approach to color description. We believe this to improve the performance gain of color description substantially. The second aim of the project is to improving the exposure of the developed techniques during the Marie Curie by extending the theory to related computer vision fields such as robotics, video retrieval and image annotation. In conclusion, the project foresees in both deepening the theory developed during the Marie Curie and improving the exposure of its techniques. This, together with the financial support for improved mobility, will much improve the candidate’s chances of a successful integration in the integration institute.
Field of science
- /natural sciences/computer and information sciences/artificial intelligence/computer vision
Call for proposal
See other projects for this call