Cord M., Cunningham P. (eds.) Machine Learning Techniques for Multimedia

Springer, 2008, -296 p.Large collections of digital multimedia data are continuously created in different fields and in many application contexts. Application domains include web searching, cultural heritage, geographic information systems, biomedicine, surveillance systems, etc. The quantity, complexity, diversity and multi-modality of these data are all exponentially growing. The main challenge of the next decade for researchers involved in these fields is to carry out meaningful interpretations from these raw data. Automatic classi fication, pattern recognition, information retrieval, data interpretation are all pivotal aspects of the whole problem. Processing this massive multimedia content has emerged as a key area for the application of machine learning techniques. ML techniques and algorithms can ‘add value’ by analysing these data. This is the situation with the processing of multimedia content. The ‘added value’ from ML can take a number of forms:
by providing insight into the domain from which the data are drawn,
by improving the performance of another process that is manipulating the data or
by organising the data in some way.
This book brings together some of the experience of the participants of the European Union Network of Excellence on Multimedia Understanding through Semantics Computation and Learning (www.muscle-noe.org). The objective of this network was to promote research collaboration in Europe on the use of machine learning (ML) techniques in processing multimedia data and this book presents some of the fundamental research outputs of the network.
In the MUSCLE network, there are multidisciplinary teams including expertise in machine learning, pattern recognition, artificial intelligence, information retrieval or image and video processing, text and cross-media analysis. Working together, similarities and differences or peculiarities of each data processing context clearly emerged. The possibility to bring together, to factorise many approaches, techniques and algorithms related to the machine learning framework has been very productive.
We structured this book in two parts to follow this idea: Part I introduces the machine learning principles and techniques that are used in multimedia data processing and analysis. A comprehensive review of the relevant ML techniques is first presented. With this review we have set out to cover the ML techniques that are in common use in multimedia research, choosing where possible to emphasise techniques that have sound theoretical underpinnings. Part II focuses on multimedia data processing applications, including machine learning issues in domains such as content-based image and video retrieval, biometrics, semantic labelling, human–computer interaction, and data mining in text and music documents. Most of them concern very recent research issues. A very large spectrum of applications is presented in this second part, offering a nice coverage of the most recent developments in each areaPart I Introduction to Learning Principles for Multimedia Data
Introduction to Bayesian Methods and Decision Theory
Supervised Learning
Unsupervised Learning and Clustering
Dimension Reduction
Part II Multimedia Applications
Online Content-Based Image Retrieval Using Active Learning
Conservative Learning for Object Detectors
Machine Learning Techniques for Face Analysis
Mental Search in Image Databases: Implicit Versus Explicit Content Query
Combining Textual and Visual Information for Semantic Labeling of Images and Videos
Machine Learning for Semi-structured Multimedia Documents: Application to Pornographic Filtering and Thematic Categorization
Classification and Clustering of Music for Novel Music Access Applications