A Computational Framework for Segmentation and Grouping

Rating:
Summary: Amazing new Early Vision Methodology!
Review: The book provide a brand new methodology and mature computational framework to solve early vision problems. It uses the tensor to convey the information and vote for each other based on tensor fields. After accumulating the votes on each input sites, it gives out the surface, curves and juctions all at once through feature extraction procedure. This method have found a lot of applications in motion analysis, stereo analysis, segmentation... So far, this book is the only book introducing this amazing new Tensor Voting methodology and software.

Rating:
Summary: A fusion of other papers with many errata
Review: This book covers the essential parts of the Tensor Voting framework, describes some of its applications, and compares it with other methodologies. If you want to implement this new approach, however, you're left alone with many details that are not explained, like the construction of the voting fields (especially in the 3D case) and how to choose their parameters. Another thing that annoyed me is that the book is just a fusion of the PhD theses of the two co-authors Mi-Suen Lee and Chi-Keung Tang. Almost all of its text was simply copied from the other documents, thereby introducing several errata and inconsistencies, since some passages are hard to understand without the original accompanying figures. Some of the given equations are erroneous, which can be quite misleading, and many references to other literature are specified with an incorrect number.

Although you won't find a lot of information about the Tensor Voting framework in general and this is the only book covering the topic, I wouldn't recommend buying it. It's not worth its price. You'll be better off getting the original papers -- for free!