Customer Reviews

Computer Vision: A Modern Approach

5 star:		(5)
4 star:		(1)
3 star:		(2)
2 star:		(3)
1 star:		(10)

 

 

I found this book a tough read, and I have studied quite a few computer vision books over the years. Oddly enough, the figures are very good and instructive, but the mathematics portion of the book is quite involved for someone who does not have a background in advanced mathematics, including probability theory, advanced calculus, and numerical analysis. Quite a few steps are omitted in the derivation of the equations to the point that I am not sure what many of the integrals actually mean. The algorithm boxes throughout the book are helpful, but often they overcomplicate the algorithms involved. This overcomplication of subjects starts right from the beginning, when the background computer vision material is being developed. Even such basic subjects as linear filtering, color, and edge detection are presented in an oblique and confusing manner. I would assume that the book was written mainly to present the more advanced material in some kind of novel fashion. Instead, the chapters on applications- "Finding in Digital Libraries", and "Image-Based Rendering" are largely a textual explanation with no algorithms to speak of on HOW to accomplish these complex tasks. If you are an introductory student of this topic, might I recommend "Computer Vision" by Linda Shapiro. It has all of the introductory material that this book has in a much clearer presentation. Plus, there is even some information on advanced topics such as content based image retrieval that gives clear algorithmic instructions. For more advanced students, I recommend Trucco & Verri's "Introductory Techniques for 3D Computer Vision". Trucco & Verri's book contains the algorithm boxes much like this book has, but I feel that the algorithm steps have a more concise format, and I have been able to code a number of algorithms by following their instructions. I have had no such luck with this book. Just because this book is the very latest to be published on this subject, does not mean that it is the most instructive.

First of all, the field of computer vision requires a requisite level of mathematical sophistication in order to engage constructively the many topics. These methods are not difficult, per se, but borrow from many other disciplines and simply require time and effort.

Forsyth and Ponce do not provide the necessary qualitative and intuitive descriptions of the various concepts prior to engaging in the rigorous mathematics. Moreover, they simply do a shabby job of explaining some rather simple concepts in mathematics and their application to Vision. They babble on and on about very simple concepts, over complexifying them, and then spend too little time giving intuitive overviews.

The 5 star reviews are written by the authors and/or publishers. No self respecting PhD student, Masters student, or teacher, other than the author, would ever endorse this collossal waste of money. My own Vision professor said the only reason he has it up there is because there isn't an alternative.

When receiving a poor review, authors typically respond with, "well, you need to have such and such a basis before appreciating the book". This is basically, self-admitted failure. Granted, a book like this should not go into the depths of linear algebra, and should fairly assume some basic requisities. However THIS book just flat out fails as a TEXT-BOOK. As a book to teach students computer vision, it is just horrible. Not much more needs to be said. The authors cannot clearly, logically, and intuitively explain concepts. Its as simple as that. Has nothing to do with the intelligence of the reader.

A certain well known professor in Vision called it, and I quote, "utter garbage as a student text".

This book was very confusing even for a first year grad student like me...it is not a good book to learn Computer Vision from the beginning. I had to turn to other texts to better understand the material. I was highly disappointed that my professor chose this text for first year grad class in Computer Vision. Unfortunately, I cannot recommend this book. There are other books out there that are so much better to learn Computer Vision.

This is, from what I can tell, just about the most complete up-to-date text in the field of computer vision as of late 2006.

But it's a mess.

I'm a PhD student, and have worked my way through more than my fair share of high-level computer science textbooks. This one makes me really appreciate many of them. It reads like a first draft -- overly wordy at times, skipping over important issues, poorly organized... Some concepts that ought to be really simple are made very painful due to what seems to be laziness on the part of the editor. It's like the only people that critiqued this book prior to publication already knew all there is to know about computer vision.

A particularly nasty aspect of this book is that the authors have a horrible habit of including some term in some complex mathematical formula, with no reference whatsoever to that term in the surrounding text! In an explanation of how to use Expectation Maximization in line-fitting, they include a standard-deviation term, with no mention of how you're supposed to choose a value for it other than "...for sigma as before". The only "before" in which the SD (sigma) is mentioned in a similar context that I can find is IN THE PREVIOUS CHAPTER!!!

Anyway, if you want to try to teach yourself vision, don't bother. If you need the book for a class, I'm sorry it's so expensive. Either way, don't expect much.

this book sucks! terms are often inconsistent and expressions always unclear. Many simple concepts are written in very complicated ways. Perhaps this is because the author want this book to look more "academic". Don't buy this book unless you are already an expert in this area

I have to heartily agree with the previous reviewer. This was the assigned text for my postgrad Computer Vision course also and I found it completely incomprehensible. From a practical perspective I don't see how any use can be made of this book.
In contrast Gonzalez&Wood's 'Digital Image Processing' was a joy to read.

It is hard for me, as a Ph.D. graduate student to understand this book. The content is not well organized and many terms are not defined before using.

Not recommend for any person.

Good point: probably most recent vision book
Bad point: hard to read, missing obvious computer vision topics

The Authors used too many technical words without definition, and beautified each sentence by using complicated structures. Because of that, it took so much time to read and understand than any other similar books.

This book is very difficult to read, specially mathematical notation. Authors use some strange notations very awkwad expression!