Image segmentation using Spiking Pixel Architecture

Amine Bermak , M. Hopfinger , Dominique Martinez (1)

1 CORTEX - Neuromimetic intelligence

This paper describes a spiking pixel architecture that can be used in a locally interconnected network in order to perform image capture as well as image segmentation. In this network, interaction between pixels can be locally obtained and is mediated via the discharge speed of each spiking pixel. The paper studies three different architecture of locally coupled pixels and shows that synchronisation of two coupled pixels can be obtained using only local excitation schemes. A theoretical condition for synchronisation was also deduced and which confirms the obtained simulation results. The proposed schemes were also successfully tested for a set of noisy gray scale images.

Mots clés: image sensors; neurones impulsionnels; spiking neurons; capteurs d'image

Domaines: Autre [cs.OH]

Amine Bermak, M. Hopfinger, Dominique Martinez. Image segmentation using Spiking Pixel Architecture. Special session on Next Generation Image Sensors for Multimedia applications - SCI'2002, Jul 2002, Orlando, Florida, USA, 5 p. ⟨inria-00107570⟩