Skip to main content
Fachgebiete
Automobil + Motoren Bauwesen + Immobilien Business IT + Informatik Elektrotechnik + Elektronik Energie + Nachhaltigkeit Finance + Banking Management + Führung Marketing + Vertrieb Maschinenbau + Werkstoffe Versicherung + Risiko
DE
EN
Themenseiten
Organisationspsychologie Projektmanagement Marketing Smart Manufacturing
Bücher
Zeitschriften
Weitere Formate
Podcasts Webinare Technik Webinare Wirtschaft Kongresse Veranstaltungskalender Awards
Jetzt Einzelzugang starten
Zugang für Unternehmen
Referenzkunden
MTZ-Fachkongress

Antriebe und Energiesysteme von morgen


Austausch von Automobil- und Energiewirtschaft

Am 14./15. Mai geht es in Chemnitz um die Mobilitätswende durch Elektro- und Wasserstofffahrzeuge.
Erweiterte Suche
Anmelden
nach oben

2019 | Buch

Fourier Transform Kapitel lesen Erstes Kapitel lesen

Fundamentals of Image Data Mining

Analysis, Features, Classification and Retrieval

verfasst von: Dr. Dengsheng Zhang

Verlag: Springer International Publishing

Buchreihe : Texts in Computer Science

Enthalten in: Springer Professional "Wirtschaft+Technik" , Springer Professional "Technik" , Springer Professional "Wirtschaft"

Einloggen, um Zugang zu erhalten
insite
SUCHEN

Über dieses Buch

This reader-friendly textbook presents a comprehensive review of the essentials of image data mining, and the latest cutting-edge techniques used in the field. The coverage spans all aspects of image analysis and understanding, offering deep insights into areas of feature extraction, machine learning, and image retrieval. The theoretical coverage is supported by practical mathematical models and algorithms, utilizing data from real-world examples and experiments.

Topics and features: describes the essential tools for image mining, covering Fourier transforms, Gabor filters, and contemporary wavelet transforms; reviews a varied range of state-of-the-art models, algorithms, and procedures for image mining; emphasizes how to deal with real image data for practical image mining; highlights how such features as color, texture, and shape can be mined or extracted from images for image representation; presents four powerful approaches for classifying image data, namely, Bayesian classification, Support Vector Machines, Neural Networks, and Decision Trees; discusses techniques for indexing, image ranking, and image presentation, along with image database visualization methods; provides self-test exercises with instructions or Matlab code, as well as review summaries at the end of each chapter.

This easy-to-follow work illuminates how concepts from fundamental and advanced mathematics can be applied to solve a broad range of image data mining problems encountered by students and researchers of computer science. Students of mathematics and other scientific disciplines will also benefit from the applications and solutions described in the text, together with the hands-on exercises that enable the reader to gain first-hand experience of computing.

Inhaltsverzeichnis

Frontmatter

Preliminaries

Frontmatter
Chapter 1. Fourier Transform
Abstract
This chapter presents an evolutional journey from Fourier series to Fourier transform and to discrete Fourier transform. It aims to prepare readers with mathematical fundamentals and build a foundation for understanding wavelet theories. It begins with sinusoids that are the building blocks of Fourier transform. After the fundamental properties of sinusoids are understood, Fourier series and Fourier transform are formally introduced. It is followed by describing DFT and sampling theory. The connection between DFT and its application on signal processing is systematically revealed to the readers. By finishing the Fourier transform chapter, readers will be able to understand the power and limitations of Fourier transform, this naturally leads to windowed FT in the following chapter.
Dengsheng Zhang
Chapter 2. Windowed Fourier Transform
Abstract
This chapter presents the time–frequency analysis theory by introducing two variants of Fourier transform, i.e., short-time Fourier transform and Gabor filters. The chapter first analyzes the limitation of FT in capturing time and spatial info from a signal or image. It then introduces a naïve windowed FT or STFT to overcome the limitation. A more powerful Gaussian windowed FT or Gabor transform is then introduced to further improve FT’s capability. Difference between FT and windowed FT is discussed with illustrations. By understanding how STFT and Gabor filters work, readers are prepared for the more advanced and powerful contemporary wavelet theories.
Dengsheng Zhang
Chapter 3. Wavelet Transform
Abstract
This chapter presents the multi-resolution analysis theory by formally introducing the contemporary wavelet theories. It starts with formulating a wavelet transform as a transform similar to windowed FT but at multiple resolutions or scales. It then uses the simplest wavelet i.e. Haar wavelet to demonstrate step-by-step how both 1D and 2D discrete wavelet transforms (DWT) work. A 2D wavelet decomposition tree is used to help readers understanding 2D DWT. Readers are then demonstrated with a DWT application on image analysis. By finishing this chapter, readers will have a full understanding how DWT works, what types of features a wavelet can capture and how they can be used for image data mining.
Dengsheng Zhang

Image Representation and Feature Extraction

Frontmatter
Chapter 4. Color Feature Extraction
Abstract
This chapter focuses on one of the three major types of image features; colors. It first gives a brief introduction to color science, followed by the introduction of four color spaces commonly used in image feature extraction. Readers are demonstrated with pros and cons of each color space. Two segmentation techniques are also shown to divide an image into regions. In the next, different types of histogram features are introduced to give readers ideas on how simple features can be extracted from a color image. Finally, a number of most commonly used color features are described and discussed in details including four color descriptors standardised by MPEG-7 such as CSD, DCD, CLD, and SCD. This chapter is also a shortcut to color science, which is a complex theory.
Dengsheng Zhang
Chapter 5. Texture Feature Extraction
Abstract
This chapter focuses on another image feature called the texture feature. Two types of texture feature methods are discussed: traditional spatial methods and contemporary spectral methods. The chapter first introduces four spatial or handcrafted methods including Tamura, GLCM, MRF, and FD. Readers are shown how microstructures of an image can be modeled with handcraft and statistical methods. Pros and cons of spatial methods are also discussed. In the next, four spectral texture methods are described in details including DCT, Gabor, wavelet, and curvelet. These spectral texture methods not only match theory with data but also strengthen readers’ understanding of the powerful spectral transforms introduced in Part I. Readers are demonstrated with rich working examples and illustrations.
Dengsheng Zhang
Chapter 6. Shape Representation
Abstract
This chapter covers a large number of commonly used shape methods which are broadly classified into two major categories: contour-based methods and region-based methods. It begins with introducing some simple and perceptual shape descriptors followed by several shape signatures in contour-based methods. More contour-based methods are described in the next including moments, stochastic method, scale space method, FD, and shape context. Several structural contour methods are also covered including syntactic, chain code, polygon, and smooth curve decomposition. The rest of the chapter covers region-based shape methods including geometric moments, complex moments, GFD, shape matrix, and shape profiles. Structural methods such as convex hull and media axis are also discussed. Two MPEG-7 shape descriptors are also discussed in the chapter, i.e., CSS and ART. Most of the techniques discussed in this chapter are demonstrated with illustrations.
Dengsheng Zhang

Image Classification and Annotation

Frontmatter
Chapter 7. Bayesian Classification
Abstract
This chapter covers the first of four image classification methods in the book. It begins with a brief introduction of Bayesian theorem and demonstrates how Bayesian theorem works using two examples. The rest of the chapter focuses on applications of Bayesian theorem on image classification. It begins with describing how images can be classified using a simple Naïve Bayesian classifier and its variants. The NB methods give readers the opportunity to relate image data with Bayesian theorem before contacting with other sophisticated Bayesian methods. After these preparations, three more powerful and nonparametric Bayesian methods are described in algorithmic forms and illustrations. In the next, more advanced methods such as parametric method and Gaussian process are discussed with illustrations.
Dengsheng Zhang
Chapter 8. Support Vector Machine
Abstract
This chapter covers another major and powerful machine learning tool which is SVM. The chapter begins with the introduction of linear classifier, K-NN classifier, and perceptron which are the key to understand discriminative approaches such as SVM and ANN. After these preparations, the primal form SVM is formally introduced. It then goes on to transform primal form SVM to dual form SVM with detailed explanation and justification. In the next, kernel SVM is formally defined and techniques of building new kernels are described in details. For computation purpose, the kernel trick is introduced and an application of SVM using PMK is also demonstrated. The chapter concludes with three fusion techniques to build multi-class SVM classifiers.
Dengsheng Zhang
Chapter 9. Artificial Neural Network
Abstract
This chapter covers one of the hottest topics and one of the most powerful machines in computing world, i.e., CNN. The chapter begins with the introduction of artificial neurons and a review of perceptron. A simple ANN and nonlinear ANN are formally introduced in the next. After describing activation and inhibition techniques, a very important BP algorithm is formulated and illustrated in great details. The second part of the chapter dedicates to CNN which is a breakthrough development on ANN technology. This part begins with the introduction of CNN architecture followed by step-by-step anatomy and description of the CNN convolution layers. The chapter concludes with the visualization and Matlab implementation of a real CNN architecture. Readers are shown with the insight view of the actual CNN with great details.
Dengsheng Zhang
Chapter 10. Image Annotation with Decision Tree
Abstract
This chapter introduces one of the most intuitive and traditional learning machines which is DT. The chapter begins with the introduction of DT concept and the formulation of DT using a standard ID3 algorithm with focus on splitting criterion. It is followed by describing another commonly used DT algorithm: C4.5. After the introduction of both ID3 and C4.5, the two subbranches of CART are described: classification tree and regression tree. An application of regression tree is shown on a real dataset to demonstrate how the CART actually works. The chapter is highlighted and concluded with an application of DT on image classification using a working example or case study. It demonstrates how a DT is built step by step and how the DT is used for prediction.
Dengsheng Zhang

Image Retrieval and Presentation

Frontmatter
Chapter 11. Image Indexing
Abstract
This chapter deals with image indexing, which is a technique to organize large numbers of images in the database into a data structure so that images can be properly retrieved. The chapter first demonstrates two numeric indexing methods: list indexing and tree indexing. In the next, a more powerful inverted file indexing method is introduced, beginning with textual document indexing. The description then focuses on image indexing using inverted file, specifically, how varieties of term weights are determined in image indexing.
Dengsheng Zhang
Chapter 12. Image Ranking
Abstract
This chapter deals with image ranking, a technique to determine the similarity between images in the database so that only relevant images are retrieved. A number of important and commonly used similarity measures are discussed including the Lp distance, mass-based distance, cosine distance, χ2 statistics, HI, quadratic distance, and Mahalanobis distance. The pros and cons of each of the distances are highlighted. The next part of the chapter covers performance measures such as RPP, F-measure, PWH, PSR, and Bullseye. The pros, cons, and relationships between them are also explained in details. Readers are shown how similarity measures and performance measures work together to retrieve images correctly.
Dengsheng Zhang
Chapter 13. Image Presentation
Abstract
This chapter deals with image presentation, a technique to present images to user after images are retrieved from an image database. Variety of ways of presenting images to users or visualizing an image database are demonstrated to readers in this chapter. The chapter begins with simple caption browsing to category browsing which is very common on the web. More powerful and sophisticated content browsing techniques are introduced in the next, including content browsing in 3D, with focus and force-directed content browsing. The chapter is highlighted and concluded with Google image search which is a sophisticated application of variety of image presentation techniques discussed earlier in the chapter.
Dengsheng Zhang
Backmatter
Metadaten
Titel
Fundamentals of Image Data Mining
verfasst von
Dr. Dengsheng Zhang
Copyright-Jahr
2019
Electronic ISBN
978-3-030-17989-2
Print ISBN
978-3-030-17988-5
DOI
https://doi.org/10.1007/978-3-030-17989-2

Premium Partner

    Bildnachweise