Abstract--Endoscopy is a standard imaging modality commonly used in different medical fields like lesion diagnosis in hollow organs or mini-invasive surgery. Meanwhile, endoscopic data suffer from the fact that each image of a video-sequence only corresponds to a small 2D field of view. This paper presents a mosaicing algorithm leading to visually coherent large field of view maps. The ability of the algorithm to build 2D bladder maps is assessed with both phantom and patient data. This contribution describes also a 3D cartography method and gives preliminary surface reconstruction results on phantoms with bladder textures.