In the scope of the development of IFREMERs hybrid underwater vehicle named 'HROV Ariane', seafloor mapping with optical imagery sensors is a major function which has motivated the development of a software processing tool for semi-automatic, on-line or off-line, 2D and 3D optical mapping. We present in this paper development results for a complete software framework intending to make state-of-the-art and future algorithms available for routine operation. The aim is to provide a software tool able to produce 2D/3D maps from high resolution image sequences merged with navigation data in a robust and simplified process without requiring the presence of a specialist controlling the computation process. The 2D mapping technique aims to merge many thousands of multiple view, small footprint images, into a single geo-referenced mosaic that accounts for image and navigation fusion. The 3D mapping is more suitable for smaller areas and aims to reconstruct geo-referenced and scaled 3D models of sea-floor scenes from multiple image frames. Algorithms integrated to the Matisse tool have been extensively tested on several datasets and in real world trials, ensuring robust performance in numerous environmental scenarios.