Machine Vision, Third Edition: Theory, Algorithms, Practicalities (Signal Processing and its Applications)

For the analyst wanting to get into image recognition, Davies offers a detailed look at the many methods used in the last 30-40 years. These include neural networks, support vector machines, and the Hough transform. If you are tempted to use [or are using] the OpenCV code base for image research, then the book can be a vital theoretical framework. OpenCV is about the best open source image code out there on the net, but it is poorly documented. It does come with many methods for basic and vital operations like make a grayscale image from a colour image, and making a binary image from a grayscale image. But why the code does certain things (actually many things) is rarely explained. Try using this book for understanding. Plus, the text lets you get an idea of how to modify OpenCV for your purposes. And if you are going to use this book with OpenCV, look closely at the section on using multiple classifiers for training and then testing against unknown images. It is the basic idea for the cascading classifiers used by OpenCV. Along these lines, one improvement for a future edition of the book could be an analysis of code packages that are currently available for image processing. Just a thought. But it would greatly help people wanting an expert assessment on the efficacies of available packages. Or, on a more basic level, it would aid simply in delineating what is out there.