New course added to Coursera Computer Vision.
Summary of the course
While vision is a basic task for humans, it is very complicated to endow machines with the capability of seeing. Since the 1960s, this area of research has tremendously evolved and new insights have been gained. However, many questions still remain open or are not satisfactorily answered, yet.
In this course, we will get familiar with the formal language and basic algorithms of computer vision, recalling the necessary math whenever it is needed. We start with low-level image processing such as edge detection, which is then used for higher-level tasks like feature extraction and correspondence estimation.
Using those correspondences together with the knowledge about how images are formated and how 3D objects move in space, we will understand the relation between stereo images and a 3D scene in terms of the Epipolar geometry. We will discuss how this concept depends on the views being calibrated or uncalibrated and will understand the concept of homographies. Using all that, we will ultimately understand the entire procedure of how a 3D scene can be reconstructed from a pair of stereo images.