Online Robust Endomicroscopy Video Mosaicking Using Robot Prior

This letter discusses the development of a mosaicking algorithm for building large- and high-resolution confocal images. Due to the nature of optics and vision systems in general, there is still a dilemma between choosing a wide field-of-view and high resolution. The most accepted solution is to opt for a high-resolution optics and expand the FOV algorithmically thanks to mosaicking approaches. The study reported in this letter consists of online and real-time construction of large mosaics using individual confocal images with a micrometer resolution. These individual images are provided by a confocal laser endomicroscopy system which can grab in vivo real-time images through a minimally invasive access. The acquisition of the confocal images is achieved by moving the imaging probe on the studied sample surface with a constant contact between the probe and the sample. The mosaicking algorithm proposed in this letter deals with the combination of both the robot inputs and the image registrations. The proposed method has demonstrated very promising performances in terms of accuracy and robustness with regard to image noise (poor image quality or loss of contact between the probe and the sample) as well as misregistration issues. Experiments carried out with a highly accurate robotic system and a ground truth obtained by conventional optical microscopy demonstrate the robustness of the proposed approach.