High dynamic range image

Night shot of New York created by exposure series

Bamberg Gate at night in Kronach . Exposure series from three images.

HDR image, composed of 3 images

Under High Dynamic Range Image ( HDRI , HDR image , "image with a high dynamic range ") or high-contrast image comprises various techniques for recording and playback of images with large differences in brightness from about 1: 1000th Classic images are referred to as low dynamic range images or LDR images if they are to be delimited to HDR .

Nowadays, HDR images can be recorded directly by very good cameras or special cameras , generated from exposure series of photos with normal dynamic range ( low dynamic range , LDR) or calculated directly as 3D computer graphics . They can only be shown directly to a limited extent on conventional (TFT) screens and media and / or in ambient light - they have to be reduced in brightness contrast for display. This process is called dynamic compression ( English tone mapping ). Regardless of this limitation, HDR images can be used to avoid overexposure and underexposure , preserve image details better and carry out more extensive image processing . Not only photography and computer graphics, but also applications such as medicine or virtual reality benefit from these advantages.

The representation of HDR photographs ranges from very natural and inconspicuous representations to impressionistic or artifact-rich artistic photographs with exaggerated colors and unmistakable halos.

principle

False color representation of the brightness stored in a conventional LDR image (left) and an HDR image (right). While the LDR image only includes 256 different brightness levels, the values ​​for the HDR image range from 0.18 to 560. The HDR image particularly takes into account details in bright areas that appear saturated in the LDR image.

Comparison of a single shot with a compact camera (left) and an HDR image after tone mapping (right). The image sections on the right have been dynamically compressed individually.

Most digital images using only 256 levels (8  bits ) for each of the red, green and blue - the color channels . This color depth is often not sufficient to reproduce the differences in brightness that occur in natural scenes. Higher color depths are usually rarely used, since screens and print media are not capable of displaying them anyway.

The surroundings visible from a camera or a viewer typically have a dynamic range (ratio of the highest and lowest luminance ) on the order of 10,000: 1. The dynamic range is even greater when a light source is directly visible or both an interior and an exterior area illuminated by sunlight can be seen. Human visual perception is able to adapt to lighting conditions that range over nearly ten orders of magnitude (a factor of 10 ¹⁰ ); up to approximately five orders of magnitude are visible at the same time within a scene. In photography it is also common to specify the dynamic range in light values (LW). Another unit for specifying the dynamic range is the decibel (dB).

In contrast to visual perception, photographs that were created with conventional digital cameras often suffer from overexposure and underexposure. With high dynamic range imaging, image files are generated with a dynamic range that can better capture the brightnesses occurring in nature in their entirety. The pixel values ​​are proportional to the actual luminance. Only when an HDR image is displayed is its brightness range suitably reduced. Even if almost all screens still have a low brightness range, HDR images offer advantages; For example, based on HDR images, details are retained in very dark and light areas.

History and Applications

The physically based image synthesis (“rendering”) was perhaps the first application of HDR images. The rendering software Radiance , developed by Greg Ward Larson from 1985 onwards, used floating point numbers internally to store brightness values . To be able to save the rendered images without losing brightness information, Ward developed the Radiance HDR format . Even Paul Debevec dealt with HDR techniques when he for the simulation of motion blur at a computer animation abspeicherte moving highlights with high dynamic range. As early as 1968 Oppenheim and others published the first tone mapping operator in a different context; the principles presented there have been rediscovered by some newer operators.

The applications of high dynamic range imaging cover the following areas:

• Image synthesis . With physically based renderers, brightness values ​​must be represented with a high dynamic range in order tocorrectly calculatethe interaction of light and materials within a high-contrast 3D scene . In applications such as architecturalsimulation,absolute radiometric valuesmust alsobe calculated in order to correctly assess lighting conditions. It is also possible to have 3D scenes illuminated by HDR images ( image-based lighting ). Newer graphics cards can handle rudimentary HDRI in real time, which is particularlyinterestingfor computer games .

• Digital photography . In digital photography, HDR recordings avoid overexposure and underexposure and allow an unproblematic software-based white balance . Manufacturers have increased the dynamic range of their image sensors over time, but the dynamic range of HDR images has so far only been achieved by a few special cameras. Additional work is required to generate HDR images with conventional cameras.

Application of a motion blur filter, left on an LDR image, right on an HDR image (result after tone mapping). By storing the correct brightness information in the HDR image, the result looks more lifelike, similar to an actual camera shake.

• Image editing . Newer versions of some image editing programs can edit HDR images directly. This allows brightness, contrast and color changes to be made without losses in the form of saturated pixel values. Effects and filters such as soft focus appear more realistic based on HDR images, especially with highlights.

• Digital cinema and video. The projection of digital cinema films with a medium dynamic range is foreseeable; However, such films will be produced and edited in an HDR format. MPEG-4 compatible encodings for HDR video data have already been proposed. See also high dynamic range video

• Virtual reality . Conventional panoramic images and virtual environments that areloaded and explored interactivelyvia the World Wide Web suffer particularly from overexposure and underexposure. If panoramas are available as HDR images, the image sections visible at a certain point in time can be dynamically compressed individually, which creates a more natural image effect.

• Surveillance systems and machine vision . The ability to simultaneously display image details inside and outside of buildings is particularly useful for surveillance cameras. The robustness of HDR cameras against extreme lighting conditions is also an advantage when it comes to machine vision.

• Medicine. In endoscopy, there is a need for image sensors that are as small as possible and that deliver high-quality images even at low brightness. For example, a prototype of a sensor developed as part of the European IVP project delivers a dynamic range of over 100 dB with the smallest dimensions  . For ophthalmology be artificial retinas developed that stimulate the photoreceptor cells of the retina with visually impaired people and also have a high dynamic range.

• Architecture and lighting design . Even without photometric calibration, HDRIs provide a sufficiently precise image of the light distribution in a scene. By photographing true-to-scale architectural models, quantitative statements can be made about the brightness conditions in a planned building.

storage

Coding

There are two common ways to encode the pixel values ​​in HDR images independently of the device. Ideally, HDR codes approximate the non-linear sensory impression of the eye on brightness. This avoids that different brightness levels in the image are coded with apparently different precision and that there are visible color gradations.

One possibility is logarithmic coding, which quantizes brightnesses according to the following formula :

${\displaystyle I_{\mathrm {out} }=I_{\mathrm {min} }\left({\frac {I_{\mathrm {max} }}{I_{\mathrm {min} }}}\right)^{v}}$

Here, v is the normalized, coded brightness, which can assume values ​​between 0 and 1. Successive values ​​in this logarithmic coding have a constant ratio of , where N is the number of quantization steps. ${\displaystyle \textstyle ({\frac {I_{\mathrm {max} }}{I_{\mathrm {min} }}})^{1/N}}$

HDR data can also be encoded using a combination of mantissa and exponent (as floating point numbers). Successive floating point numbers do not have a constant ratio, but follow a sawtooth-like sequence. So that the color quantization remains invisible, the relative gradation (difference between two successive brightness values ​​divided by the value) must not exceed 1%.

Further properties of HDR encodings in addition to the relative gradation are the color space and the color depth . Storage of the absolute luminance (in the unit cd / m²) is often not done.

Formats

A JPEG HDR image that is displayed here with dynamic compression; the file contains HDR data

A distinction must be made between the coding of the pixel values ​​and the graphic format used , which determines the additional data structures in which the actual image data is embedded. Some HDR formats support multiple encodings.

The HDR formats supported by most programs are compressed without loss . However, an extension of the JPEG format was also developed for lossy storage with small file sizes. This "JPEG-HDR" image saves a dynamically compressed version of an HDR image as an ordinary JFIF file, but adds a ratio image in an additional marker that encodes the HDR information. Like other JPEG formats for HDR images, such as Kodak's ERI-JPEG or the FJPEG used by the Panorama Tools software, JPEG-HDR is currently (2009) not widely used.

The manufacturer -dependent , internal file formats of digital cameras ( raw data formats or raw formats) with 10 to 14 bits (linearly encoded, usually with offset, 1: 1,024 to 1: 15,360) offer a dynamic similar to normal sRGB 8-bit images (not linearly coded, 1: 3,300), but by no means achieve the dynamic range of HDR images. They can also be referred to as LDR formats or at best as Medium Dynamic Range, formats with a medium dynamic range . Regardless, it is possible to convert raw files to HDR formats . The scRGB coding, published as an IEC standard, also only has a mediocre dynamic range.

The following table provides an overview of various well-known HDR formats and the encodings used in them.

format	Coding	Compression	Color space	Format (bits per pixel)	Dynamic range (f-stops)	Relative gradation
Radiance HDR (.hdr, .pic)	RGBE	RLE	RGB	3 × 8 bit linear RGB, 8 bit exponent	000256	0.39 ... 0.78%
	XYZE	RLE	(CIE) XYZ	3 × 8 bit linear XYZ, 8 bit exponent	000256	0.39 ... 0.78%
TIFF (.tif, .tiff)	LogLuv24 (L logarithmic)	no	LogLuv	10 bit brightness, 14 bit chrominance signal	000016	1.09%
	LogLuv32 (L logarithmic)	RLE	LogLuv	15 + 1 bit brightness, 2 × 8 bit chrominance signal	000128	0.27%
	Floating point	no	RGB	3 × 32 bit float RGB	000254+	0.000.006 ... 0.000.012%
Portable Float Map (.pfm, .pbm)	Floating point	no	RGB	3 × 32 bit float RGB	000254+	0.000.006 ... 0.000.012%
OpenEXR (.exr)	Floating point	Wavelet , ZIP , RLE and others	RGB , (CIE) XYZ	3 × 16 bit float	000030+	0.049 ... 0.098%
SMPTE 2084 (Dolby HDR)	non-linear	none up to H.265	RGB	3 × 12 bit non-linear, PQ based EOTF	approx. 31.3	0.23% (10 4 cd / m²) 0.48% (1 cd / m²) 2.28% (0.01 cd / m²)
Hybrid Log Gamma (HDR-HLG 709, HDR-HLG 2020)	non-linear	none, JPEG up to H.265	RGB	3 × 10 bit non-linear, composite EOTF
Non-HDR: sRGB	non-linear	none, JPEG up to H.265	RGB	3 × 8 bit non-linear, gamma EOTF	000011.7	0.89% (at 100%) 3.83% (at 3%) 10% (at 0.3%)
				3 × 10 bit non-linear, gamma EOTF	000013.7	0.22% (at 100%) 0.96% (at 3%) 2.5% (at 0.3%)
				3 × 12 bit non-linear, gamma EOTF	000015.7	0.055% (at 100%) 0.24% (at 3%) 0.63% (at 0.3%)

Image generation

HDR images can be generated in three different ways: by direct recording with special cameras, indirectly from a series of differently exposed LDR images or as artificial computer graphics.

HDR cameras

Digital image sensors with a high dynamic range are under development. While some of these products are already on the market, few comprehensive solutions are available. The price for professional HDR cameras is in the range of 50,000  US dollars (2008). However, even high-quality image sensors are not yet able to completely cover the dynamic range of any natural scenes, especially of outdoor shots on a sunny day. Products marketed or developed include:

• Grass Valley's Viper Filmstream film camera is three orders of magnitude and has a dynamic range of at least ten times that of conventional digital cameras. This camera requires high-capacity hard drives to store the video data, which limits mobile use.

• SMaL Camera Technologies and Pixim have produced CMOS sensors with VGA-like resolution that can capture images with a dynamic range of four orders of magnitude at video speed. Some manufacturers offer cameras and surveillance cameras with these sensors.

• Point Gray Research has developed the LadyBug camera ("ladybug") which, thanks to several sensors, can photograph 75% of the sphere simultaneously and with a dynamic range of over four orders of magnitude.

• SpheronVR offers the SpheroCam HDR , which contains a line CCD. The automatic rotation of the camera enables panoramic images with a dynamic range of 5.5 orders of magnitude in high resolution.

• The Civetta from Weiss AG offers an automatic recording process for fully spherical HDR recordings and generates images that are calibrated using measurement technology with a dynamic range of up to 28 f-stops.
• The panorama system piXplorer 500 from CLAUSS generates metrologically calibrated HDR panorama recordings with a resolution of 500 MPixel and a dynamic range of 26 f-stops (extended version: up to 36 f-stops).

• At the Institute for Microelectronics Stuttgart , a sensor called HDRC was developed that exceeds the human eye in sensitivity and dynamic range. The effective dynamic range is 120 dB at 50 frames per second. This sensor is part of a developer kit offered by Omron .

• For professional photography, some manufacturers offer cameras with a high dynamic range, such as the Leica S1 Alpha / Pro , Jenoptik eyelike MF and LEAF c-most . These cameras were developed to encourage professional photographers who prefer films to digital images because of their higher dynamic range to switch to digital photography. The sensors of these cameras are actively cooled to suppress image noise.

• Fujifilm developed the Super CCD sensor , which contains photo diodes with different light sensitivity. An image with extended dynamic range can be created in one shot by combining the signals from the sensitive and the less sensitive diodes. The sensor has a lower effective image resolution than conventional image sensors. In addition, only the maximum detectable luminance is increased; Contrasts in shadow areas are not improved. The sensor is built into several semi-professional cameras from the Finepix series .

In addition to these products for the direct recording of HDR images, there are cameras for the amateur and semi-professional market that can automatically generate HDR or dynamically compressed LDR images from multiple images with different exposure settings (see next section). A conventional image sensor is sufficient for this. In March 2009 , the Ricoh CX1 was the first compact camera to offer this function in the form of a double exposure mode to "increase the dynamic range".

Generation from exposure series

→ Main article : HDRI generation from exposure series

With a little effort, it is possible to generate HDR images using conventional digital cameras. A series of exposures is recorded from the scene in which each image region is correctly exposed in at least one of the individual images. The individual images are then combined into an HDR image using software. It is important that the subject does not move between the individual shots. Although it is possible to correct for camera shake to a certain extent, the use of a photo tripod is recommended.

In order for the correct brightness data to be calculated from the exposure series, the light values ​​of the individual images (often saved in the Exif entries of the image files anyway ) and the opto-electronic transmission function of the camera are required. Since the transfer function is not published by most manufacturers, it should be determined yourself, ideally using a calibration scene with as many gray tones as possible. After generating the HDR image, the lens flare of the camera should be filtered out to avoid excessive light scattering in the image.

An HDR image is generated from a series of exposures (in the middle a false color display, on the right the image according to tone mapping). Thanks to HDRI, details in light and dark image regions can be recognized in equal measure, without annoying overexposure or underexposure.

A particular problem with photographic techniques is the correct recording of the directly visible or reflected sun, since massive overexposures occur even with the smallest aperture and exposure time. The correct luminance of the sun can be determined with the help of a neutral density filter or indirectly by different illuminations of a diffusely reflecting sphere.

HDR images can also be reconstructed from transparencies such as slides, negatives and film strips by multiple scanning with different exposures (see e.g. Multi-Exposure ).

HDR rendering

An image rendered using image-based lighting. The background is an HDR image that surrounds the 3D scene and illuminates the artificially modeled objects in the foreground.

→ Main article : High Dynamic Range Rendering

Newer graphics cards support hardware-based real - time rendering with a high dynamic range, often called High Dynamic Range Rendering (HDRR). This is particularly useful in computer games where the player often switches between dark and light scenes. Graphic effects such as lens scattering also look more realistic with HDRR. The achievable precision and the dynamic range are limited by the computing power available.

Image-based lighting (IBL) is an important technique in image synthesis . A scene is completely enveloped by an HDR environment map (also known as a light probe ). The easiest way to record environment maps is to use rotating cameras with fisheye lenses . Alternatively, a sphere reflecting the surroundings can be photographed or several individual photos can be combined using stitching . In order to combine IBL with well-known rendering processes such as Monte Carlo ray tracing , a special scanning strategy of the environment map is necessary so that the image noise remains low. IBL can not only be used to illuminate artificial scenes with complex light sources, but also to add artificial objects to real (film) scenes. IBL is now supported by all major 3D renderers.

presentation

HDR output devices

Diffuse reflective prints are principally LDR, as the maximum brightness depends on the ambient lighting. In order to display HDR images in print media, light emitting paper would have to be invented. At best, it would be conceivable to add glare effects in order to create the illusion of a brighter light than can be represented by the medium, as in painting. It is also possible to record printed images with a camera and then project them back onto the image (see Superimposing Dynamic Range ). Although transparencies and photographic films have a dynamic range (possibly up to ten times) greater than prints, they are problematic to use.

From a technical point of view, cathode ray tube screens have a high dynamic range because they are able to display very low, imperceptible brightnesses. In practice, however, this is irrelevant since their maximum luminance is too low for HDR images to be displayed with the desired effect. Conventional liquid crystal screens, on the other hand, are able to display high levels of brightness, but the light scattering into neighboring pixels is quite high, which limits the effective dynamic range.

The first prototypes of HDR display devices have existed since 2004 at the latest. This includes the DR37-P HDR screen from BrightSide Technologies (formerly Sunnybrook Technologies, now taken over by Dolby ). This screen is a liquid crystal screen (LCD) that is not illuminated by a uniform light source, but by a matrix of light-emitting diodes with individually adjustable brightness. Image details are displayed on the LC screen, while the large differences in brightness are modulated by the light emitting diodes. The light-emitting diode matrix can have a low resolution, since differences in brightness in the vicinity of bright pixels are masked anyway by the point spreading function of the eye. The screen brightness ranges from 0.015 to 3000 cd / m²; the contrast ratio is around 200,000: 1.

Further developments in HDR output devices can be found primarily in the digital cinema sector. Most digital projection systems for cinemas are based on the Digital Micromirror Device from Texas Instruments , a micromirror actuator . This is a high-resolution matrix of electronically controlled mirrors that can reflect light either onto a screen or onto an absorber. Gradations in brightness are created by pulse width modulation . The practical dynamic range of commercial micromirror actuators is around 500: 1.

Tone mapping

→ Main article : Tone Mapping

An HDR image after applying four different tone mapping operators

Under Tone Mapping, and Tone Reproduction called, refers to the conversion of an HDR image to an LDR image by the contrast ratio is reduced. This is necessary in order to be able to approximate an HDR image on a conventional display device or medium. The lifelike impression of brightness is lost. It is all the more important to retain the special properties of the HDR image, such as the richness of detail in dark and light image regions, as well as possible. Tone mapping operators are usually designed to produce results that are as natural as possible or rich in detail. However, some HDR software also contains operators that give the user artistic freedom.

There are different types of tone mapping operators. The simplest methods process each pixel independently. These global tone mapping operators are comparatively fast and are therefore suitable for applications in which the tone mapping has to take place in real time. So-called local or frequency-based operators are able to compress images with a particularly large contrast range without excessive loss of detail. Here, image regions with high contrast are compressed strongly, regions with low contrast less strongly. Such processes require special techniques in order to avoid image artifacts such as halos. Finally, there are gradient-based methods that weaken the brightness gradients of the HDR image.

The fact that many tone mapping operators are based on knowledge about visual perception is due not least to the fact that humans themselves seem to solve the tone mapping problem effortlessly. For example, operators can simulate brightness-dependent color and sharpness perception, which leads to more realistic results, especially with night scenes. The newer iCAM06 model takes into account a variety of effects of human perception. Many tone mapping operators require absolute brightness values.

Aesthetic aspects

Example of an artistic application of tone mapping

Example of an artistic application of tone mapping (pseudo-HDR from an image without bracketing)

One problem with the display of HDR images are halo artifacts, which often arise during tone mapping with simple local tone mapping algorithms. Modern tone mapping operators avoid such artifacts; Physiologically based operators such as iCAM06 deliver plausible results even in difficult lighting conditions.

Some HDR programs contain tone mapping operators that intentionally allow the user a great deal of freedom in setting parameters. Erik Reinhard criticizes that this would lead the user to abuse tone mapping as an effect medium. Halos, strange contrasts and overly saturated colors, which actually result from inadequacies in the tone mapping algorithm used, would be misunderstood by some users as artistic effects. This would give the wrong impression that HDRI is associated with a certain “style”. Christian Bloch encourages the creative use of tone mapping operators, but recommends calling the result “impressionistic photography” or “ hyperrealism ”, but not misleading “HDRI”.

Differentiation from exposure blending

Under the terms Exposure Blending , Exposure Fusion, "Dynamic Range Increase" or "Pseudo-HDR", methods were presented that combine differently exposed images exclusively by image processing in order to avoid overexposed and underexposed areas. HDRI techniques can be used for the same purpose by generating an HDR image from the individual images, which is then converted into an LDR image using tone mapping. However, exposure blending techniques have nothing to do with HDRI as they do not process any HDR data. Ideally, the quality of the images generated by exposure blending is comparable to that of the HDRI process.

software

→ See also: List of HDRI software under HDR software

HDR images are supported to varying degrees by full-fledged image editing programs. Adobe Photoshop from version CS 2 supports the import / export as well as the generation of HDR images, but will only offer support for some painting tools and filters in subsequent versions. The open source CinePaint , a revised version of the GIMP for cinema production , can also handle HDR images.

There are also programs that specialize in displaying, generating or tone mapping HDR images. The best known are the commercial applications FDRTools Advanced and Photomatix , the freeware programs Picturenaut , Photosphere and FDRTools Basic as well as the free software Luminance HDR .

literature

• Christian Bloch: The HDRI manual. Dpunkt, Heidelberg 2008, ISBN 3-89864-430-8
• Jürgen Held: HDR photography. The comprehensive manual. 4th edition, Rheinwerk Verlag, Bonn 2015, ISBN 978-3-8362-3012-4
• Jürgen Kircher: DRI and HDR - the perfect picture. 1st edition, Redline Verlag, 2008, ISBN 978-3-8266-5903-4
• Bernd Hoefflinger (Ed.): High-Dynamic-Range (HDR) Vision (= Springer Series in Advanced Microelectronics 26). Springer, Berlin 2007, ISBN 978-3-540-44432-9
• Axel Jacobs: High Dynamic Range Imaging and its Application in Building Research. Advances in Building Energy Research 1, 1 (2007): 177–202, ISSN  1751-2549 ( PDF, 1.5 MB ( Memento from December 3, 2008 in the Internet Archive ))
• Erik Reinhard u. a .: High Dynamic Range Imaging. Morgan Kaufman, San Francisco 2006, ISBN 0-12-585263-0

Web links

Commons : Dynamic Compressed HDR Images  - collection of images

Note: this gallery may also contain images that were created using simple exposure blending without HDR data being processed.

• Contrast bag of tricks: HDR photography (Heise Foto)
• Digital photos with increased contrast range - basics and technology of HDRI
• Art of HDR (fxguide) - History and uses of HDRI
• WebHDR - Information and web applications for HDRI (English)
• High Dynamic Range Image Encodings (Greg Ward, Anyhere Software) - Comparison of different HDR encodings
• Demonstration of an HDR display (English)
• SMPTE ST 2094 and Dynamic Metadata

Individual evidence

1. ↑ Reinhard et al: High Dynamic Range Imaging, p. 7
2. ↑ Reinhard et al: High Dynamic Range Imaging, p. 187
3. ↑ J. A Ferwada: Elements of Early Vision for Computer Graphics. IEEE Computer Graphics and Applications 21, 5 (2001): 22-33, ISSN  0272-1716 . Quoted in Reinhard u. a .: High Dynamic Range Imaging, p. 6
4. ↑ Reinhard et al: High Dynamic Range Imaging, p. 87
5. ↑ Bloch: Das HDRI-Handbuch, p. 44
6. ↑ fxguide - Art of HDR ( Memento from November 11, 2007 in the Internet Archive ), accessed on February 22, 2009
7. ^ A. Oppenheim et al.: Nonlinear Filtering of Multiplied and Convolved Signals. In: IEEE Transactions on Audio and Electroacoustics 16, 3 (September 1968), pp. 437-466, ISSN  0018-9278
8. ↑ Reinhard u. a .: High Dynamic Range Imaging, pp. 326-331
9. ↑ Reinhard et al.: High Dynamic Range Imaging, pp. 87-89
10. ↑ Reinhard u. a .: High Dynamic Range Imaging, p. 48 f.
11. ↑ Reinhard et al: High Dynamic Range Imaging, p. 88
12. ↑ R. Mantiuk include: Perception-motivated High Dynamic Range Video Encoding. ACM Transactions on Graphics 23, 3 (2004), pp. 733-741, ISSN  0730-0301
13. ↑ Reinhard et al: High Dynamic Range Imaging, p. 12
14. ↑ Hoefflinger: High-Dynamic-Range (HDR) Vision, p. 138
15. ^ Jacobs: High Dynamic Range Imaging and its Application in Building Research
16. ↑ G. Wyszecki, WS Stiles: Color Science: Concepts and Methods, Quantitative Data and Formulas. John Wiley and Sons, New York 2000. Quoted in Reinhard et al. a .: High Dynamic Range Imaging, p. 90
17. ^ G. Ward, M. Simmons: Subband Encoding of High Dynamic Range Imagery. In First ACM Symposium on Applied Perception in Graphics and Visualization (APGV), pp. 83-90. ACM, New York 2004
18. ↑ Konrad Kabaja: Storing of High Dynamic Range Images in JPEG / JFIF Files. In Proceedings of the Central European Seminar on Computer Graphics 2005 ( PDF, 3.9 MB )
19. ↑ Bloch: Das HDRI-Handbuch, pp. 52–58
20. ↑ Reinhard u. a .: High Dynamic Range Imaging, p. 13
21. ↑ Bloch: Das HDRI-Handbuch, S. 35
22. ↑ IEC 61966-2-2 (2003): Multimedia systems and equipment - Color measurement and management - Part 2-2: Color management - Extended RGB color space - scRGB.
23. ↑ Reinhard et al.: High Dynamic Range Imaging, pp. 160-164
24. ↑ Bloch: Das HDRI-Handbuch, S. 97
25. ↑ Hoefflinger: High-Dynamic-Range (HDR) Vision, p. 2
26. ↑ Bloch: Das HDRI-Handbuch, p. 103
27. ↑ http://digitalleben.t-online.de/erste-kompaktkamera-mit-integrierter-hdr/id_17771738/index
28. ↑ Reinhard et al: High Dynamic Range Imaging, p. 145
29. ^ J. Stumpfel et al .: Direct HDR Capture of the Sun and Sky. In AFRIGRAPH 2004 Proceedings, pp. 145-149. ACM, New York 2004, ISBN 1-58113-863-6 ( Online ( Memento of February 27, 2009 in the Internet Archive ))
30. ↑ Reinhard et al: High Dynamic Range Imaging, pp. 396-401
31. ^ Greg Spencer et al .: Physically-Based Glare Effects for Digital Images. In ACM SIGGRAPH 1995 Proceedings, pp. 325-334. ACM, New York 1995, ISBN 0-89791-701-4 ( PDF, 2.8 MB )
32. ^ O. Bimber et al: Superimposing Dynamic Range. ACM SIGGRAPH Asia 2008 Papers, Article No. 150, ISSN  0730-0301 ( PDF, 35 MB )
33. ↑ Reinhard et al .: High Dynamic Range Imaging, p. 171
34. ↑ Reinhard et al.: High Dynamic Range Imaging, pp. 176-179
35. ↑ Reinhard et al: High Dynamic Range Imaging, p. 9
36. ^ Jacobs: High Dynamic Range Imaging and its Application in Building Research
37. ↑ ^{a b} Reinhard et al: High Dynamic Range Imaging, p. 182 f.
38. ↑ Reinhard et al: High Dynamic Range Imaging, p. 17
39. ^ Gregory Ward Larson: A Visibility Matching Tone Reproduction Operator for High Dynamic Range Scenes. IEEE Transactions on Visualization and Computer Graphics 3, 4 (Oct.-Dec. 1997): 291–306 ( PDF, 880 kB )
40. ↑ Jiangtao Kuang et al: iCAM06: A refined image appearance model for HDR image rendering. Journal of Visual Communication and Image Representation 18, 5 (Oct. 2007), pp. 406-414, ISSN  1047-3203 ( online )
41. ↑ Erik Reinhard: Flickr HDR ( memento August 25, 2008 in the Internet Archive ), accessed on February 22, 2009
42. ↑ Bloch: Das HDRI-Handbuch, S. 189 f.
43. ↑ See for example Tom Mertens, Jan Kautz, Frank van Reeth: Exposure Fusion. In Proceedings of Pacific Graphics 2007, pp. 382-390. IEEE, Piscataway (NJ) 2007, ISBN 0-7695-3009-5 ( Online ( Memento of October 31, 2011 in the Internet Archive ))

This version was added to the list of articles worth reading on March 8, 2009 .