Chapter 1, Introduction to Computer Vision, lays out the basics of the computer vision science—what it is; where it is used; the definition of an image and its basic properties, such as pixels, depth, and channels; and so on. This short chapter is a totally introductory chapter meant for people who are completely new to the computer vision world.
Chapter 2, Getting Started with OpenCV, introduces the OpenCV library and detail its core by going through the most important building blocks of OpenCV development. You'll also be presented with information about where to get it and how to use it. This chapter will even briefly go through the usage of CMake and how to create and build OpenCV projects, after which you'll learn about the Mat class and its variants, reading and writing images and videos, and accessing cameras (among other input source types).
Chapter 3, Array and Matrix Operations, covers the fundamental algorithms that are used to create or alter matrices. In this chapter, you'll learn how to perform matrix operations, such as the cross product, the dot product, and inversion. This chapter will introduce you to many of the so-called per-element matrix operations, along with mathematical operations such as mean, sum, and Fourier transformation.
Chapter 4, Drawing, Filtering, and Transformation, covers the wide category of image-processing algorithms as much as possible within the scope of this book. This chapter will teaches you how to draw shapes and text on images. You'll learn how to draw lines, arrows, rectangles, and more. This chapter will also present you with a wide range of algorithms that are used for image-filtering operations, such as smoothing filters, dilation, erosion, and morphological operations on images. By the end of this chapter, you'll be familiar with the powerful remapping algorithm and the usage of color maps in computer vision.
Chapter 5, Back-Projection and Histograms, introduces the concept of histograms and teach you how they are calculated from single- and multi-channel images. You'll learn about the visualization of histograms for grayscale and color images, or, in other words, histograms calculated from the hue values of pixels. In this chapter, you'll also learn about back-projection images; that is, the reverse operation of histogram extraction. Histogram comparison and equalization are also among the topics covered in this chapter.
Chapter 6, Video Analysis – Motion Detection and Tracking, explains how to process videos, especially for operations such as real-time object detection and tracking, using some of the most popular tracking algorithms in computer vision. After a brief introduction to how to process videos in general, you'll learn about the Mean Shift and CAM Shift algorithms, along with Kalman filtering, using real-world examples and object-tracking scenarios. By the end of this chapter, you'll also have learned about background- and foreground-extraction algorithms and how they are used in practice.
Chapter 7, Object Detection – Features and Descriptors, starts with a brief introduction to object detection using template matching, and then moves on to teach you about a wide range of algorithms, that can be used for shape analysis. The topics covered in this chapter also include chain of keypoint detection, descriptor extraction, and descriptor matching, which are used for object detection based on features instead of simple pixel colors or intensity values.
Chapter 8, Machine Learning in Computer Vision, covers the Machine Learning (ML) and Deep Neural Network (DNN) modules of OpenCV and some its most important algorithms, classes, and functions. Starting with the SVM algorithm, you'll learn how to train a model based on groups of similar training and then use that model to classify input data. You'll learn how to use HOG descriptors with SVM to classify images. This chapter also covers the implementation of artificial neural networks in OpenCV, and then moves on to teaches you about cascade classification. The last section of this chapter will teach you how to use pretrained models from third-party libraries, such as TensorFlow, to detect multiple objects of different type in real-time.