This thesis contributes to the domain of Human Computer Interaction (HCI). It is to propose to investigate solutions that enable blind computer users to have access of digital pictures from multimedia interfaces. The central concern of this work relies on the creation of a non-visual transformation model allowing the blind user to build a satisfactory mental image. By satisfactory we mean that the encoding must allow the characterization of both the semantic and morphologic properties. By taking into account several precepts of multimodal human computer interaction, the proposed solution has been to work conjointly with the audio and haptic media. Results from our investigations on the non-visual encoding of graphical data revealed that the modalities steming from the auditory media seem to be better adapted to convey semantic attributes. Haptic interaction on the other hand seems to be useful to symbolize morphologie and spatial attributes. Based on our findings, we have developed IMAGIN (Interface for Multimodal Access to Graphical INformation), an audio-haptic system that enables blind computer users to explore digital pictures using the hearing and feeling modalities. This system relies on the "edition <-> exploration" concept in which a sight specialist firstly defined the image into audio and haptic form. A blind person can then subsequently explore this audio-haptic image without any external help. We conclude this dissertation by purposing an audio and haptic extension to a traditional image format.