Image segmentation is a fundamental task in image processing, computer vision, and associated pattern recognition research efforts. This book, as stated by editor Yu-Jin Zhang, attempts to bring together a selection of state-of-the-art work on the segmentation of images and image sequences.

First, the introductory chapter (Section I) gives an overview of the collective published efforts in the last 40 years on image segmentation analysis, from the basic definitions to the statistics of the task.

The following chapters are organized according to the three levels of segmentation research, namely: algorithm development, algorithm evaluation, and the systematic study of evaluation methods.

Sections II and III comprise case studies on image and video segmentation algorithms respectively. As we know, segmentation could be equated with boundary definition and estimation. The techniques to achieve this goal generally make use of the uniformity and continuity characteristics of the image or image sequence. First, F. Escolano and M. Lozano present three case studies of optimal image segmentation using dynamic energy minimization. It is followed by the work by G. Bellettini and R. March on variational models for image segmentation to recover a piecewise smooth approximation of an image together with a discontinuity set. A graph- and meta-heuristic-based technique is also found here as well as a technique that has originated from chaotic non-linear dynamic modelling to disambiguate the pixel clustering in image segmentation.

Mean-shift and content-based techniques are presented next. I. Gu and V. Gui utilize joint space-time-range mean-shift with success for both image and video. It is followed by automatic methods based on scene changes and active shape models that are of great practical interest.

Section IV consists of algorithm descriptions for segmenting particular types of images, e.g. multi-channel, texture, and medical images. S. Dai and Y. Zhang have proposed a two-step feature and image space segmentation framework for segmenting color images.  It is followed by a texture-classification technique using variography to select the optimal neighbourhood window size. Next, 3D medical image segmentation and multi-dimensional morphology based segmentation studies are discussed.

Section V concentrates on applying image segmentation techniques using domain specific knowledge to solve ambiguities and enhance the segmented output. These applications include automatic lip segmentation from color image or video, restoration and segmentation of degraded character images for OCR, segmentation of food images, and application of segmentation for blind navigation.

Finally,  Section VI discusses aspects of objectively evaluating segmentation algorithms using supervised and unsupervised techniques.  As one would imagine for an image analysis book, there are many line drawings and gray-scale image examples before and after the processing algorithms. This is good, however many are a bit small to view and color is not used.

This book is very rich in content due to the fact that it collects about 20 high-quality contributions from many well-known authors in this field. Each chapter is a good introduction to the particular technique discussed, and each contains a useful list of reference publications for further reading. In my opinion, this book is a handy reference for researchers and graduate students in computer science and engineering who are working on images. It is also valuable for academics and professionals who are interested in the mathematical modelling of images and video sequences. Finally, it is useful for practitioners for applying image segmentation in medical studies, the food industry, and information processing.

BOOKSBOOKSBOOKS

Advances in

Image and Video Segmentation

by

Yu-Jin Zhang, Editor

IRM Press, 2006

Reviewed by:  Kai Huang, Silverbrook Research