Computer Vision: A Modern Approach (2nd Edition)

Very helpful and gives a modern approach for sure. Very good gift for a student or anyone interested in computers and going deeper into them.

Terrible book. I bought this book because my graduate class in computer vision required it, but I had hard time understanding what the author is trying to say. I think the author expects you to have prior experience in computer vision and throws a bunch of jargons without explaining much. I ended up relying only on the lectures and skipped reading this book, and I did well in the class.

This book is the required textbook for a class. I ordered the 2nd edition after reading the poor reviews dating back to well before the second edition had been published. It was my hope that perhaps a major rewrite had occurred after that author absorbed the poor reviews. Sadly, that does not appear to be the case. I am just starting the class and have much of the book left to read, but the frustration felt by past reviewers is now clearly understood by me.In general, the book tends to run both hot and cold. Clearly the authors are deeply knowledgeable about the subject. And there are many good sections of the book that do reasonable job of explaining the concepts. It is the sections that dive into the math where the authors tend to undermine their own hard work and that, I believe, are source of a number of the book's poor reviews.One of the topics of the book is the classification (of images), so in keeping with this let me attempt to classify this book. This book seems to fall into the category of a book written by really smart people that loses more than it has to when read by a not-so-smart person. Said differently, this is probably a great book for those that already know the subject and the math. That is, the book can be quite readable for a few paragraphs and give the reader some intuition for the topic. However, when the book dives into the math for a few paragraphs and then emerges on the far shore, it has probably abandoned most readers. Certainly, more often than not it has abandoned me and, from the reviews, I believe that I am not the only one. I suspect the authors have majors in math, but perhaps they have been swimming in the math for so long that it has become comfortable. Whatever the reason, the authors do a poor job in setting up the math, are too dependent on concise math notation (again which would be well know to a math major but not necessarily by others) and as the math discussion progresses the authors make among the largest leaps mathematically among any textbook that I can recall. Additionally, the sections that discuss the math seem to be of varying quality. Often I would get the sense that what made it to final print were a set of notes or a rough draft that had not yet been completed let alone edited.Regarding the poor setup and excessive brevity in a passage I just read, the authors describe an equation as having three terms, then they give the equation and never state which term is which. As the math unfolds, the reader is stuck because, although he can read the transformations relating to (for example) the rho term he does not know which term is represented by rho. The author also at points use what I find to be a somewhat odd notation (e.g., for linear algebra). In general a few examples would go a long way as well. The authors attempts at homogeneous coordinates and scale rotate and translate was awful and awfully short. A book that I have on motion in robotics spends perhaps 5 times as many pages to discuss a similar topic, with more examples and less dense notation, with the upshot being that the reader of the robotics book walks away having learned the concept and the Computer Vision reader walks away confused and having learned little (again, a math major might have been able to track better with the CV book and gotten a little more).Were the authors to read this critique, they might counter that they were trying limit the math and instead to provide narratives that provide for an intuitive understanding of the subject. This would be a fair point and a worthy goal. When writing a book, the author must decide who his intended audience is; e.g., an undergraduate engineering student, a graduate student, academic colleagues, etc. However, if this was their intent then the approach the authors took was not unlike trying to meet a page count by removing every third word from the draft text.In short, of what I have read so far, this book can only be followed without great struggle by those already familiar with the subject and the math. It has lots of potential, so I hope that the authors take the criticism in the constructive spirit in which it is offered with the result being a still better third edition.

had to get this for a college class and it was worthless. the authors cannot explain complex topics to save their lives. stick with your professor's lectures and skip this book.

I bought this without reading the fine print (shame on me). Apparently, the "kindle" version doesn't actually work on the Kindle. Nor does it work on any android device except the Kindle Fire. It does say it works on "PC", by which it means Windows only... This is an unacceptable delivery method. It has a relatively large number of errors. The material is presented in an unapproachable way, such that it's probably most useful as a reference for someone who just needs a reminder of available techniques. I used this for a semester course in graduate computer vision. I'm also a math major, so my problem with the book is not a lack of proper background.

Read its first 5 pages and found it's hard to understand. Will update it after reading further.

Don't bother buying. It is a crappy print on demand edition with a hardcover pasted over it. Find it free online from the authors and print it yourself cheaper and better than the publisher.

I am a very senior computer vision professor at a top-10 university. I never post on Amazon, but this book is simply awful. How can this fellow be editor in chief of PAMI? This is not how computer vision should be taught. It's more a book for computer graphics people, and not a good one.