A Neuronal Planar Modeling for Handwriting Signature based on Automatic Segmentation

This paper deals with offline handwriting signature verification. We propose a planar neuronal model of signature image. Planar models are generally based on delimiting homogenous zones of images; we propose in this paper an automatic segmentation approach into bands of signature images. Signature image is modeled by a planar neuronal model with horizontal secondary models and a vertical principal model. The proposed method has been tested on two databases. The first is the one we have collected; it includes 6000 signatures corresponding to 60 writers. The second is the public GPDS-300 database including 16200 signature corresponding to 300 persons. The achieved results are promising.