"Le choix d'Alice" (Alice's choice)
Shape Recognition and Visual Strategy

n° 385 - July 2000


Scientists working in the field of artificial vision are intrigued by the visual strategies the human eye appears to adopt. Even when researchers use artificial retinas with optical properties that are at least the equivalent of the human eye, they have great difficulty reproducing the eye's ability to recognize shapes.

Saccadic eye movements "interrogate" the scenes upon which we cast our eyes. The duration of each saccade (between one tenth of a second and a half second) as well as the fixation of one's gaze on an image appear to be optimized according to the visual strategy and the task to be performed, with different strategies being adopted for different tasks-for example, the memorization of a scene, the identification of an object in a picture, or the reading of a printed text.

French scientists at the "Laboratoire de statistique et probabilités" (Laboratory of Statistics and Probabilities, CNRS-University of Lille 1) and the "Centre de mathématiques et de leurs applications" (Center for Mathematics and Applications, CNRS-Ecole Normale Supérieure in Cachan) are currently investigating the modeling of artificial vision. They are studying visual strategies using mathematics and probability games such as the one known as "choix d'Alice" (Alice's choice), pattern analysis, etc. This field offers examples of simple situations that, surprisingly, defy mathematical and statistical strategies. When probability is applied to artificial vision, the results are capable of accelerating algorithms-for example, for the tracking of roads in satellite images or for face detection.

References :

  • Model-Based Classification Trees. Donald German, Bruno Jedynak. Forthcoming in IEEE Transactions on Information Theory.

  • An Active Testing Model for Tracking Roads in Satellite Images. Donald German, Bruno Jedynak. IEEE Transactions on Pattern Analysis and Machine Intelligence. January 1996.

  • Efficient Focusing and Face Detection. Yali Amit, Donald German, and Bruno Jedynak. Face recognition: From theory to applications, H. Wechsler et al., NATO Asi Series, F. Springer Verlag, Berlin, 1997.

    For more information, see the following web site:


  • Previous page

    CNRS online - © CNRS URL : http://www.cnrs.fr URL in the US : http://www.cnrs.org