In this chapter, we began discussion features and its importance in computer vision applications. Harris Corner Detector is used to detect corners where runtime is of utmost importance. These can run on embedded devices with high speeds. Extending over to more complex detectors, we saw FAST features and in combination with BRIEF descriptors, ORB features can be formed. These are robust for different scales as well as rotations. Finally, we saw the application of feature matching using ORB features and a use of pyramid downsampling.
The discussion on black box features will continue in the next chapter with the introduction of neural networks and especially CNNs.
