This chapter describes technology for automatic and robust detection of patterns in realistic image processing scenarios, developed in the framework of the IM2 NCCR. The images come from a wearable head-mounted camera, the WearCam, and are subjected to severe perturbations, including drastic changes in lighting conditions and motion blur. Pattern recognition is used in two capacities: to detect the direction of gaze of the user and to detect specific elements in the visual field of the user (faces and other objects). Applications of these methods for early detection of autism spectrum disorders (ASD) are briefly presented.