Principles of Digital Image Processing, Volume 3: Advanced Methods (Undergraduate Topics in Computer Science)

Principles of Digital Image Processing, Volume 3: Advanced Methods (Undergraduate Topics in Computer Science)

Wilhelm Burger, Mark J. Burge

Language: English

Pages: 374

ISBN: 2:00260318

Format: PDF / Kindle (mobi) / ePub


This easy-to-follow textbook is the third of three volumes which provide a modern, algorithmic introduction to digital image processing, designed to be used both by learners desiring a firm foundation on which to build, and practitioners in search of critical analysis and concrete implementations of the most important techniques. This volume builds upon the introductory material presented in the first two volumes (Fundamental Techniques and Core Algorithms) with additional key concepts and methods in image processing.

Features and topics:
* Practical examples and carefully constructed chapter-ending exercises drawn from the authors' years of experience teaching this material
* Real implementations, concise mathematical notation, and precise algorithmic descriptions designed for programmers and practitioners
* Easily adaptable Java code and completely worked-out examples for easy inclusion in existing (and rapid prototyping of new) applications
* Uses ImageJ, the image processing system developed, maintained, and freely distributed by the U.S. National Institutes of Health (NIH)
* Provides a supplementary website with the complete Java source code, test images, and corrections—www.imagingbook.com
* Additional presentation tools for instructors including a complete set of figures, tables, and mathematical elements

This thorough, reader-friendly text will equip undergraduates with a deeper understanding of the topic and will be invaluable for further developing knowledge via self-study.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

v). Additional examples are shown in Fig. 2.14. Note that the selected radius r is obviously too small for the structures in images Fig. 2.14 (c, d), which are thus not segmented cleanly. Better results can be expected with a larger radius. With the intent to improve upon Niblack’s method, particularly for thresholding deteriorated text images, Sauvola and Pietikäinen [116] proposed setting the threshold to Q(u, v) = µR (u, v) · 1 + κ · σR (u, v) −1 σmax , (2.66) with κ = 0.5 and σmax = 128

filter (b) follows the same construction principle but consists of five subregions whose side length is odd [127]. In both types of filters, all subregions have the same size; r = 1 (a) and r = 2 (b), respectively. • • • • • • • • • • • • • • • R1 • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • R2 • • • • • R9 • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • R3 • • • • • R8 • • • • • • • • • • • • • • • • • • • •

do 25: for j ← d− , . . . , d+ do 26: a ← I(u+i, v+j) 27: S1 ← S1 + a ◃ Eqn. (5.4) 28: S2 ← S2 + a 2 ◃ Eqn. (5.5) 29: s ← S2 − S12 /n ◃ subregion variance (s ≡ n · σ 2 ) 30: µ ← S1 /n ◃ subregion mean (µ) 31: return (s, µ). 5.1 Kuwahara-type filters 125 Algorithm 5.3 Color version of the Kuwahara-type filter (adapted from Alg. 5.1). The algorithm uses the definition in Eqn. (5.11) for the total variance σ2 in the subregion R (see line 24). The vector µ (calculated in line 25) is the average

nonlinear R′ , G′ , B ′ components), or the lightness component L of the CIELAB and CIELUV color spaces (see Sec. 3.1 for a detailed discussion). Algorithm 5.7 can be easily adapted to implement this type of color filter. An obvious disadvantage of this method is that it naturally blurs across color 5.3 Anisotropic diffusion filters 153 edges if the neighboring colors are of similar brightness, as the examples in Fig. 5.18 (e–g) demonstrate. This limits its usefulness for practical

stands for a particular characteristic of the entire shape. If one component is changed, the whole shape will change. The advantage is that it is possible to capture coarse shape properties with only a few numeric values, and the level of detail can be increased (or decreased) by adding (or removing) descriptor elements. In the following, we describe what is called “cartesian” (or “elliptical”) Fourier descriptors, how they can be used to model the shape of closed 2D contours and how they can be

Download sample

Download