By Petra Perner (eds.)
Despite being a tender box of study and improvement, facts mining has proved to be a winning method of extracting wisdom from large collections of based electronic info assortment as often saved in databases. while info mining was once performed in early days totally on numerical info, these days multimedia and net functions force the necessity to improve information mining tools and methods which could paintings on every kind of knowledge similar to files, photographs, and signals.
This booklet introduces the elemental thoughts of mining multimedia info and demonstrates easy methods to follow those equipment in a variety of software fields. it truly is written for college students, ambitioned pros from and drugs, and for scientists who are looking to give a contribution R&D paintings to the sector or follow this new technology.
Read or Download Data Mining on Multimedia Data PDF
Similar imaging systems books
The second one variation of a bestseller, this e-book is a pragmatic advisor to snapshot processing for the ordinary and technical sciences neighborhood. scholars, practitioners, and researchers can achieve instant entry to a legitimate simple wisdom of picture processing by way of referencing basic ideas within the usual sciences.
Self-contained textual content protecting useful photo processing tools and idea for snapshot texture research. ideas for the research of texture in electronic photos are necessary to a number functions in parts as varied as robotics, defence, medication and the geo-sciences. In organic imaginative and prescient, texture is a vital cue permitting people to discriminate gadgets.
The world of content-based video retrieval is a really sizzling zone either for examine and for advertisement functions. as a way to layout powerful video databases for functions similar to electronic libraries, video construction, and a number of net functions, there's a nice have to improve powerful suggestions for content-based video retrieval.
This book addresses disbursed embedded shrewdpermanent cameras –cameras that practice on board research and collaborate with different cameras. This e-book presents the fabric required to higher comprehend the architectural layout demanding situations of embedded shrewdpermanent digicam structures, the hardware/software environment, the layout method for and purposes of dispensed clever cameras including the state of the art algorithms.
- WDM Technologies: Passive Optical Components (WDM Technologies)
- Principles of Digital Image Processing: Advanced Methods
- Digital Video Transcoding for Transmission and Storage
- Nuclear Technology Review
Extra resources for Data Mining on Multimedia Data
2 The Case-Based Reasoning Process The CBR process is comprised of six phases (see Figure 36): • • • • • • Current problem description Problem indexing Retrieval of similar cases Evaluation of candidate cases Modification of a selected case, if necessary Application to a current problem: human action. The current problem is described by some keywords, attributes or any abstraction that allow to describe the basic properties of a case. Based on this description indexing of case base is done. Among a set of similar cases retrieved from the case base the closest case is evaluated as a candidate case.
5 Discretization of Attribute Values A numerical attribute may take any value on a continuous scale between its minimal value x1 and its maximal value x2. Branching on all these distinct attribute values does not lead to any generalization and would make the tree very sensitive to noise. Rather we should find meaningful partitions on the numerical values into intervals. The intervals should abstract the data in such a way that they cover the range of attribute values belonging to one class and that they separate them from those belonging to other classes.
Cj x1j x2j . . xij . . xnj Rj ... ... ... ... ... ... Cm SUM x1m L1 x2m L2 . . . xim Li . . . 2 Gini Function This measure takes into account the impurity of the class distribution. The Gini function is defined as: m G = 1 − ∑ pi2 i =1 The selection criteria is defined as: IF Gini( A) = G(C) − G(C / A) = Max! 5 Discretization of Attribute Values A numerical attribute may take any value on a continuous scale between its minimal value x1 and its maximal value x2. Branching on all these distinct attribute values does not lead to any generalization and would make the tree very sensitive to noise.
Data Mining on Multimedia Data by Petra Perner (eds.)