Bull Communicating Pictures

A Course in Image and Video Coding
1. Auflage 2014
ISBN: 978-0-08-099374-4
Verlag: Elsevier Science & Techn.
Format: EPUB
Kopierschutz: 6 - ePub Watermark

E-Book, Englisch, 560 Seiten

Communicating Pictures starts with a unique historical perspective of the role of images in communications and then builds on this to explain the applications and requirements of a modern video coding system. It draws on the author's extensive academic and professional experience of signal processing and video coding to deliver a text that is algorithmically rigorous, yet accessible, relevant to modern standards, and practical. It offers a thorough grounding in visual perception, and demonstrates how modern image and video compression methods can be designed in order to meet the rate-quality performance levels demanded by today's applications, networks and users.With this book you will learn: - Practical issues when implementing a codec, such as picture boundary extension and complexity reduction, with particular emphasis on efficient algorithms for transforms, motion estimators and error resilience - Conflicts between conventional video compression, based on variable length coding and spatiotemporal prediction, and the requirements for error resilient transmission - How to assess the quality of coded images and video content, both through subjective trials and by using perceptually optimised objective metrics - Features, operation and performance of the state-of-the-art High Efficiency Video Coding (HEVC) standard - Covers the basics of video communications and includes a strong grounding in how we perceive images and video, and how we can exploit redundancy to reduce bitrate and improve rate distortion performance - Gives deep insight into the pitfalls associated with the transmission of real-time video over networks (wireless and fixed) - Uses the state-of- the-art video coding standard (H.264/AVC) as a basis for algorithm development in the context of block based compression - Insight into future video coding standards such as the new ISO/ITU High Efficiency Video Coding (HEVC) initiative, which extends and generalizes the H.264/AVC approach

Professor David R. Bull PhD, FIET, FIEEE, CEng. obtained his PhD from the University of Cardiff in 1988. He currently holds the Chair in Signal Processing at the University of Bristol where he is head of the Visual Information Laboratory and Director of Bristol Vision Institute, a group of some 150 researchers in vision science, spanning engineering, psychology, biology, medicine and the creative arts. In 1996 David helped to establish the UK DTI Virtual Centre of Excellence in Digital Broadcasting and Multimedia Technology and was one of its Directors from 1997-2000. He has also advised Government through membership of the UK Foresight Panel, DSAC and the HEFCE Research Evaluation Framework. He is also now Director of the UK Government's new MyWorld Strength in Places programme. David has worked widely across image and video processing focused on streaming, broadcast and wireless applications. He has published over 600 academic papers, various articles and 4 books and has given numerous invited/keynote lectures and tutorials. He has also received awards including the IEE Ambrose Fleming Premium for his work on Primitive Operator Digital Filters and a best Paper Award for his work on Link Adaptation for Video Transmission. David's work has been exploited commercially and he has acted as a consultant for companies and governments across the globe. In 2001, he co-founded ProVision Communication Technologies Ltd., who launched the world's first robust multi-source wireless HD sender for consumer use. His recent award-winning and pioneering work on perceptual video compression using deep learning, has produced world-leading rate-quality performance.

Autoren/Hrsg.

Bull, David

Leseproben

List of figures
Fig. 1.1 A geometric interpretation of compression. 3 Fig. 1.2 The multimedia communications jigsaw puzzle. 7 Fig. 1.3 Simplified high level video compression architecture. 11 Fig. 1.4 The scope of standardization. 13 Fig. 1.5 A chronology of video coding standards from 1990 to the present date. 14 Fig. 2.1 The visible spectrum. 20 Fig. 2.2 Cross-section of the human eye. (Public domain: http://commons.wikimedia.org/wiki/File:Schematic_diagram_of_the_human_eye_en.svg.) 21 Fig. 2.3 Fundus image of a healthy retina. (Public domain from: http://commons.wikimedia.org/wiki/File:Fundus_photograph_of_normal_right_eye.jpg.) 22 Fig. 2.4 The focal length of the lens. 23 Fig. 2.5 Photoreceptor distribution in the retina. (Reproduced with permission from: Mustafia et al. [12].) 24 Fig. 2.6 Normalized rod and cone responses for the human visual system. (Reproduced with permission from: Bowmaker and Dartnall [33]. //www.ncbi.nlm.nih.gov/pmc/articles/PMC1279132/. Avail Wikimedia Commons.) 25 Fig. 2.7 Retinal cell architecture (Public domain image adapted from http://commons.wikimedia.org/wiki/File:Retina-diagram.svg). 27 Fig. 2.8 Spatial opponency, showing center-surround on cell and its firing pattern due to excitation. 28 Fig. 2.9 The visual cortex. (Reproduced from: http://www.expertsmind.com/topic/neuroscience/eye-and-visual-pathways-93024.aspx.) 28 Fig. 2.10 Mach band effect. 33 Fig. 2.11 Adelson’s grid. (Reproduced with permission from: http://web.mit.edu/persci/people/adelson/checkershadow_illusion.html.) 33 Fig. 2.12 CIE luminous efficiency curve. (Public domain image: http://en.wikipedia.org/wiki/File:CIE_1931_Luminosity.png.) 34 Fig. 2.13 Dark adaptation of rods and cones. 35 Fig. 2.14 Increased immersion from color images. 36 Fig. 2.15 Opponent processing of color. 38 Fig. 2.16 Color dependence on context. The bottom picture is just a squashed version of the top one yet the green stripes become blue. 38 Fig. 2.17 The CIE 1931 chromaticity chart. (Reproduced with permission from Ref. [21].) 39 Fig. 2.18 Just-noticeable differences at different contrast increments. 40 Fig. 2.19 JND curve for human vision. 41 Fig. 2.20 Contrast sensitivity chart. 41 Fig. 2.21 Luminance and chrominance CSF responses. 42 Fig. 2.22 Luminance contrast sensitivity function. 43 Fig. 2.23 Texture change blindness. (Images courtesy of Tom Troscianko.) 45 Fig. 2.24 The importance of phase information in visual perception. Left: original. Right: phase distorted version using the complex wavelet transform (Reproduced with permission from Vilankar et al. [14]). 46 Fig. 2.25 Perspective-based depth cues can be very compelling and misleading. 47 Fig. 2.26 Pits and bumps—deceptive depth from lighting. 48 Fig. 2.27 The hollow mask illusion. 49 Fig. 2.28 Spatio-temporal CSF. (Adapted from Kelly [18].) 51 Fig. 2.29 Variation of critical flicker frequency (Reproduced with permission from Tyler [23]). 52 Fig. 2.30 Eye movements in response to task (Public domain image from: http://commons.wikimedia.org/wiki/File:Yarbus_The_Visitor.jpg). 54 Fig. 2.31 Example of texture masking. 54 Fig. 2.32 Edge masking for high and low dynamic range content. 55 Fig. 2.33 Temporal masking effects for various edge step sizes. (Reproduced with permission from Girod [30].) 55 Fig. 3.1 Spectral characteristics of sampling and aliasing. 66 Fig. 3.2 Demonstration of aliasing for a 1-D signal. Top: sinusoid sampled below Nyquist frequency. Bottom: Fourier plot showing spectral aliasing. 67 Fig. 3.3 2-D spectral characteristics of sampling and aliasing. Left: Top—original signal spectrum; Bottom—sampled signal spectrum with no aliasing. Right: Top—original signal spectrum; Bottom—sampled signal spectrum with aliasing due to sub-Nyquist sampling. 68 Fig. 3.4 Hexagonal sampling lattice and its reciprocal as defined by equation (3.8). 69 Fig. 3.5 Example image histogram for 256 × 256 image Stampe_SV4. 70 Fig. 3.6 Autocorrelation plots for Acer image (512 × 512). Top left to bottom right: original image; autocorrelation function for row 100; autocorrelation function for whole image; 2-D autocorrelation surface. 72 Fig. 3.7 Autocorrelation plots for Stampe_SV4 image (512 × 512). Top left to bottom right: original image; autocorrelation function for row 100; autocorrelation function for whole image; 2-D autocorrelation surface. 73 Fig. 3.8 Autocorrelation plots for Stampe_SV4 image (256 × 256). Top left to bottom right: original image; autocorrelation function for row 100; autocorrelation function for whole image; 2-D autocorrelation surface. 74 Fig. 3.9 Temporal autocorrelation plots for Foreman (30 fps). Top to bottom right: sample frame showing selected 16 × 16 block used; temporal correlation for a single pixel; temporal correlation for a 16 × 16 block. 75 Fig. 3.10 Filterbank responses for the LeGall low-pass and high-pass analysis filters. 80 Fig. 3.11 Filter response for H.264 half-pixel interpolation filter. 81 Fig. 3.12 Common uniform quantizer characteristics. 85 Fig. 3.13 Common non-uniform quantizers. Left: center deadzone. Right: Lloyd Max quantizer. 86 Fig. 3.14 Feedforward linear prediction. Top: encoder. Bottom: decoder. 88 Fig. 3.15 Prediction signal dynamic range. Top: input signal. Bottom left: distribution of 1000 samples of input signal. Bottom right: distribution of 1000 samples of prediction residual. 89 Fig. 3.16 Feedback-based linear prediction. 90 Fig. 3.17 Feedback-based linear predictor with quantization noise modeling. 91 Fig. 3.18 Self-information and probability. Left: plot of self-information vs probability for a single event. Right: plot of the self-information of an event weighted by its probability. 95 Fig. 4.1 Image sample array. 100 Fig. 4.2 Image samples. 101 Fig. 4.3 Pixelation at varying resolutions. Top left to bottom right: 256 × 256; 64 × 64; 32 × 32; 16 × 16. 102 Fig. 4.4 Typical macroblock structure. 103 Fig. 4.5 Typical group of pictures structure and prediction modes. 105 Fig. 4.6 Aspect ratios of common formats, normalized according to resolution. 107 Fig. 4.7 Widescreen formats. 108 Fig. 4.8 Variation of field of view with viewing distance (aspect ratio = 16:9 here). 109 Fig. 4.9 Interlaced vs progressive frame scanning. 111 Fig. 4.10 Example of effects of interlaced scanning with poor...