The aim of this course is to take a closer look at the image acquisition process, taking into account two characteristics of the scenes imaged:
They evolve over time. Thus, the first part of this course will focus on video, which makes it possible to acquire this temporal aspect. We will look at motion estimation techniques within videos (optical flow, block matching methods with different search strategies and parametric motion - in addition, deep learning methods will be discussed), followed by automatic object tracking methods (Kalman and particle filtering);
Before being projected onto the sensor in two dimensions, they have three spatial dimensions. Thus, calibration and stereo-vision techniques (enabling 3D information to be retrieved from two photographs of a scene) will be studied. Finally, the use of 3D point clouds of the scene and the reconstruction of surfaces from them will be discussed.