Contrast estimation for the intensity images. A new approach with a dynamic and architectural modelisation of the PLE layer retina

François Devillard (1) , Bernard Heit (1) , Gérard Brunelot (1)

1 CRAN - Centre de Recherche en Automatique de Nancy

Currently, the knowledges of the mammal visual systems allow the design of new methods for computer vision. These methods try to benefit some properties observed in the visual system. In this paper, our research works use some known mechanisms of the retinian perception. We have designed a artificial retina that implements the first layers dedicated to the primary contrast perception. The contrast detection and its estimation are very important because they are the first steps to a lot of high level treatments in image processing. We have oriented our study toward biological retina to better understand how works the visual system to perceive image contrasts with different lighting conditions. Our work has been focused on the functionnings discovered in the external plexiform layer (PLE). This layer integrates some interesting adaptations that improves its robustness and its efficiency in case of noisy images. The proposed algorithm is original because it can produce some adaptative effects described in the retina. The adaptations are set locally according to the conditions of lighting. The designed model using an array of filters is able to adjust locally the processing. Morever the filtering is processing temporally continuously to gain in robustness and efficiency. The suggested architecture allows too an real time use for embedded implementations based on digital signal processors or wired operators.

Mots clés: Computer vision; Retina modelling

Domaines: Traitement du signal et de l'image [eess.SP]; Traitement du signal et de l'image [eess.SP]

François Devillard, Bernard Heit, Gérard Brunelot. Contrast estimation for the intensity images. A new approach with a dynamic and architectural modelisation of the PLE layer retina. Jan 2003, pp.69-72. ⟨hal-00119044⟩