In the last few chapters, we have discussed the extraction of an object and semantic information from images. We saw how good feature extraction leads to object detection, segmentation, and tracking. This information explicitly requires the geometry of the scene; in several applications, knowing the exact geometry of a scene plays a vital role.
In this chapter, we will see a discussion leading to the three-dimensional aspects of an image. Here, we will begin by using a simple camera model to understand how pixel values and real-world points are linked correspondingly. Later, we will study methods for computing depth from images and also methods of computing the motion of a camera from a sequence of images.
We will cover the following topics in the chapter:
- RGDB dataset
- Applications to extract features from images
- Image formation
- Aligning of images
- Visual odometry...
