1:30pm - 4:30pm, Monday, July 2nd, 2007
The Future of Digital Photography and Video
http://www.viplab.cs.nott.ac.uk/icme2007
The sequence of photographs on the left were taken by a
conventional camera with
exposure values ranging from 1/1000 ~ 1/4. The
quality of the photographs is poor and non of them can show the full
detail of the scene because there are not enough bits to represent the
full dynamic range of the scene. Photographs
courtesy of Raanan Fattal, Dani Lischinski, Michael Werman, The Hebrew
University of Jerusalem.
The image on the right is rendered from a high dynamic range radiance map constructed using this sequence of low dynamic range photographs. To view the full resolution image visit this link: http://www.viplab.cs.nott.ac.uk/research4.html
The image on the right is rendered from a high dynamic range radiance map constructed using this sequence of low dynamic range photographs. To view the full resolution image visit this link: http://www.viplab.cs.nott.ac.uk/research4.html
A Brief
Overview
Throughout the history of imaging science, there has been constant quest for more advanced technologies to provide better and more real visual experiences for the consumers. The invention of colour has forever changed the imaging industry and the introduction of digital technology has made imaging ubiquitous. Now a new imaging technology has started to emerge which has the promise of bringing a new revolution in digital imaging. This new technology is called high dynamic range imaging, or HDRI for short.
Almost all of today’s image and video file formats, e.g., jpeg, mpeg, etc, use 8 bits per colour channel. This means that they can record a luminance dynamic range of about 2 orders of magnitude. Yet, the real world scenes humans experience every day have far higher luminance dynamic ranges. For instance, a scene showing both shadows and sunlit areas will have a dynamic range exceeding 100,000:1.
Human visual system is capable of simultaneously perceiving light intensities over a range of 4 orders of magnitude, and with adaptation, its sensitivity can stretch to 10 orders of magnitude.
Therefore, it is clear that current image file formats are poor representations of real scenes and only record a fraction of the contrast that humans are capable of perceiving.
Also, conventional computer monitors and other reproduction media such as printing papers have limited dynamic ranges often less than 2 orders of magnitude. This is why traditionally there is little reason to represent images with more than 8 bits per colour channel. On the image capture side a similar argument can be made. Most cameras limit their outputs to eight bits per colour channel.
However, the situation is slowly changing. In recent years, a new research area generally referred to as high dynamic range imaging (HDRI) has been gathering momentum. In HDRI, the image files record the actual colour and dynamic range of the original scene rather than the limited gamut and dynamic range of the monitor or other reproduction media. This means that image processing, manipulation, display and other operations will no longer be limited by the number of bits used to represent each pixel.
HDRI will have widespread applications in the field of multimedia including, digital cinema, digital photography, computer games, etc., and will open up many new possibilities, including dramatically improving the visual realism of digital photography and video.
As current technology can already produce high enough spatial resolution but lacks dynamic range of the pixels, many have advocated that the future of digital photography and video will be high dynamic range imaging. It has been argued that the significance of the transition from low dynamic range imaging to high dynamic range imaging can be compared to that of the transition from black and white television to colour television. In the next decades, the imaging industry will inevitably move to HDRI which will affect all aspects of imaging research, including capturing (sensor, camera), storage (compression, coding) and reproduction (rendering, tone-mapping, printing and display).
Content Sketch
The purpose of this tutorial is to provide an introduction to high dynamic range imaging. Emphasis will be on the image processing perspective of HDRI technology. The tutorial will begin by discussing the weaknesses of current (low dynamic range) image and video formats and the limitations of conventional reproduction media such as video monitors. It will then introduce the potential of high dynamic range imaging. Subsequently, it will introduce topics in all stages of the high dynamic range imaging pipeline including, capturing, processing, coding and compression, and displaying. It will conclude by discussing potential applications of HDRI technology. Indicative topics include:
(i) Introduction
- Limitations of current digital imaging systems – will include an explanation of why current digital cameras often produce poor quality photographs, especially for high contrast scenes.
- Fundamentals of human visual system – will include a brief discussion of the ability of the HVS to simultaneously perceive high contrast scenes and other fundamental properties of the HVS.
- HDRI and its potential – will include a discussion of how HDRI technology can overcome the difficulty of current technology and the technological challenges of HDRI.
(ii) HDR image and video acquisition
- Introduction to HDR image and video acquisition issues – will include an introduction to the basics of digital camera; and explanation of the fundamental concepts such as camera response function.
- Multiple exposure techniques – will include discussion of the technique of Debevec and Malik.
- High dynamic range cameras – will include discussion of the technique of Mitsunaga and Nayar.
- High dynamic range video capture – will include an introduction to the technique of Kang et al.
(iii) HDR image processing
- Introduction to image processing issues in HDR image and video
- HDR image and video file formats
- Lossless coding of HDR image and video – will include RGBE, SGI LogLuv TIFF, and others
- Lossy coding of HDR image and video – will include an introduction to techniques by Xu et al and Mantiuk et al
- Backward compatible high dynamic range MPEG/JPEG compression – will introduce the techniques of Ward et al and Mantiuk et al
- HDR image compression quality measures
(iv) HDR image display/reproduction
- Introduction to HDR image and video reproduction issues – will include an introduction to the fundamentals of CRT/LCD monitors.
- Tone mapping for HDR image and video – will discuss different approaches to tone mapping HDR images
- Global tone reproduction curve based techniques – will include discussion of techniques by Ward et al and Qiu et al.
- Local tone reproduction operator based techniques – will include discussion of techniques by Durand et al, Fattal et al, and Li et al.
- Hardware techniques and equipment for the display of HDR image and video – will include discussion of the technique of Seetzen et al and an introduction to Brightside HDR monitors.
(v) Applications of HDRI and the future of digital photography and video
- Discussion of potential applications of HDRI in digital photography, computer games, scientific and medical imaging, and other areas.
The Lecturer
Guoping Qiu is a Reader (Associate Professor) in Visual Information Processing in the School of Computer Science, the University of Nottingham, UK. Before joining Nottingham in 2000, he was a Lecturer (Assistant Professor) in the School of Computing, the University of Leeds, UK (1999 – 2000); and the School of Mathematics and Computing, the University of Derby, UK (1993 – 1999).
He received his BSc in Electronic Measurement and Instrumentation from the University of Electronic Science and Technology of China in 1984 and his PhD in Electronic Engineering from the University of Central Lancashire, Preston, UK in 1993. He has been performing research in fields related to image processing for more than 15 years and has authored more than 100 publications.
His research interests include high dynamic range imaging. He has been performing research in HDRI since 2002. He has developed several successful tone-mapping operators for displaying high dynamic range images. In August 2006, his paper on HDR image tone mapping using optimization received a Best Paper Award at the 18th International Conference on Pattern Recognition (ICPR2006). Apart from publishing academic papers on HDR imaging, with his industrial collaborators, he has filed a patent on HDRI. He is currently a consultant for a company helping the development of HDRI products. His research group at Nottingham is currently developing advanced tone mapping, coding and compression techniques for high dynamic range image and video data. He is a co-guest editor of a Special Issue on High Dynamic Range Imaging for the Journal of Visual Communication and Image Representation. More about his research can be found here: http://www.viplab.cs.nott.ac.uk/
