Projet LINX / LINX Project

Centre de Morphologie Mathématique

Sujet/Subject : Le texte est omniprésent dans notre quotidien et nous apporte sans cesse quantité d'informations pratiques indispensables. Le handicap visuel, qui touche plus d'un million de personnes en France, prive un pourcentage significatif de la population de ces informations importantes. L'objectif de notre projet est de développer un lecteur automatique pour aider les personnes mal-voyantes dans leur quotidien en leur rendant accessible une partie de l'information textuelle. Afin d'être réellement au service de l'utilisateur, ce lecteur devra être portable, simple d'utilisation, sans contrainte de manipulation rédhibitoire et capable de restituer l'information de facon adaptée.

The text is omnipresent and brings us essential information for our daily life. Visual impairment, which affects more than one million people in France, deprives a significant percentage of the population of this important information. The goal of our project is to develop an automatic reader to help visually impaired in their daily life, making available to them a part of the textual information. LINX system should be portable, easy to use, adapted to visual impaired people, capable of reproducing information appropriately.

CMM contributions

  1. Local blur estimation.
  2. Scene-text segmentation
  3. Character level annotated database

Downloads

LINX database contains 16 images acquired with smartphone cameras by visually impaired people. Some images are shown hereafter. In spite of the reduced number of images, an important number of words is present: about 1200 words with 4000 characters. Image contents are very heterogeneous ranging from text documents to products exposed in a supermarket, very blur text, noisy regions, shadow regions, high saturated regions, small and large texts, clear and dark texts etc. You can download the annotated database from here.

The whole database, 160 images acquired by visually impaired people, can be found here.

Publications

  1. T. Chabardès and B. Marcotegui. "Local blur estimation based on toggle mapping". Proceedings of 12th International Symposium on Mathematical Morphology, May 2015, Reikjavik, Island. paper.
  2. A. Belhedi and B. Marcotegui. Adaptive scene-text binarisation on images captured by smartphones. IET Image Processing , Volume 10, Issue 7, Pages 515--523, July 2016. paper.

Financeurs/Financers

Projet financé dans le cadre du FUI (Fond Unique Interministeriel) 14.
Ile de France  DGE  Cap Digital 

Partenaires/Partners :

epita  open_wide 
armines  itesoft  a+b  braillenet 
[Retour au CMM] [Back to CMM]