Abstract: In spite of the great advances in laparoscopic surgery, this type of surgery still shows some difficulties during its realization, mainly caused by its complex maneuvers and, above all, by the loss of the depth perception. Unlike classical open surgery --laparotomy-- where surgeons have direct contact with organs and a complete 3D perception, laparoscopy is carried out by means of specialized instruments, and a monocular camera (laparoscope) in which the 3D scene is projected into a 2D plane --image. The main goal of this thesis is to face with this loss of depth perception by making use of Simultaneous Localization and Mapping (SLAM) algorithms developed in the fields of robotics and computer vision during the last years. These algorithms allow to localize, in real time (25 $\thicksim$ 30 frames per second), a camera that moves freely inside an unknown rigid environment while, at the same time, they build a map of this environment by exploiting images gathered by that camera. These algorithms have been extensively validated both in man-made environments (buildings, rooms, ...) and in outdoor environments, showing robustness to occlusions, sudden camera motions, or clutter. This thesis tries to extend the use of these algorithms to laparoscopic surgery. Due to the intrinsic nature of internal body images (they suffer from deformations, specularities, variable illumination conditions, limited movements, ...), applying this type of algorithms to laparoscopy supposes a real challenge. Knowing the camera (laparoscope) location with respect to the scene (abdominal cavity) and the 3D map of that scene opens new interesting possibilities inside the surgical field. This knowledge enables to do augmented reality annotations directly on the laparoscopic images (e.g. alignment of preoperative 3D CT models); intracavity 3D distance measurements; or photorealistic 3D reconstructions of the abdominal cavity recovering synthetically the lost depth. These new facilities provide security and rapidity to surgical procedures without disturbing the classical procedure workflow. Hence, these tools are available inside the surgeon's armory, being the surgeon who decides to use them or not. Additionally, knowledge of the camera location with respect to the patient's abdominal cavity is fundamental for future development of robots that can operate automatically since, knowing this location, the robot will be able to localize other tools controlled by itself with respect to the patient. In detail, the contributions of this thesis are: - To demonstrate the feasibility of applying SLAM algorithms to laparoscopy showing experimentally that using robust data association is a must. - To robustify one of these algorithms, in particular the monocular EKF-SLAM algorithm, by adapting a relocalization system and improving data association with a robust matching algorithm. - To develop of a robust matching method (1-Point RANSAC algorithm). - To develop a new surgical procedure to ease the use of visual SLAM in laparoscopy. - To make an extensive validation of the robust EKF-SLAM (EKF + relocalization + 1-Point RANSAC) obtaining millimetric errors and working in real time both on simulation and real human surgeries. The selected surgery has been the ventral hernia repair. - To demonstrate the potential of these algorithms in laparoscopy: they recover synthetically the depth of the operative field which is lost by using monocular laparoscopes, enable the insertion of augmented reality annotations, and allow to perform distance measurements using only a laparoscopic tool (to define the real scale) and laparoscopic images. - To make a clinical validation showing that these algorithms allow to shorten surgical times of operations and provide more security to the surgical procedures.