Since the inception of Robotics, visual information has been incorporated in order to allow the robots to perform tasks that require an interaction with their environment, particularly when it is a changing environment. Depth perception is a most useful information for a mobile robot to navigate in its environment and interact with its surroundings. Among the different methods capable of measuring the distance to the objects in the scene, stereo vision is the most advantageous for a small, mobile robot with limited energy and computational power. Stereoscopy implies a low power consumption because it uses passive sensors and it does not require the robot to move. Furthermore, it is more robust, because it does not require a complex optic system with moving elements. On the other hand, stereo vision is computationally intensive. Objects in the scene have to be detected and matched across images. Biological sensory systems are based on simple computational elements that process information in parallel and communicate among them. Analog VLSI chips are an ideal substrate to mimic the massive parallelism and collective computation present in biological nervous systems. For mobile robotics they have the added advantage of low power consumption and high computational power, thus freeing the CPU for other tasks. This dissertation discusses two stereoscopic methods that are based on simple, parallel cal- culations requiring communication only among neighboring processing units (local communication). Algorithms with these properties are easy to implement in analog VLSI and they are also very convenient for digital systems. The first algorithm is phase-based. Disparity, i.e., the spatial shift between left and right images, is recovered as a phase shift in the spatial-frequency domain. Gábor functions are used to recover the frequency spectrum of the image because of their optimum joint spatial and spatial-frequency properties. The Gábor-based algorithm is discussed and tested on a Khepera miniature mobile robot. Two further approximations are introduced to ease the analog VLSI and digital implementations. The second stereoscopic algorithm is difference-based. Disparity is recovered by a simple calculation using the image differences and their spatial derivatives. The algorithm is simulated on a digital system and an analog VLSI implementation is proposed and discussed. The thesis concludes with the description of some tools used in this research project. A stereo vision system has been developed for the Webots mobile robotics simulator, to simplify the testing of different stereo algorithms. Similarly, two stereo vision turrets have been built for the Khepera robot.