Automated High Dynamic Range Imaging
Software & Images
Last update: May 30, 2007
Table of Contents 
High Dynamic Range Imaging

High Dynamic Range (HDR) imaging is the process of taking "photographs" (usually referred to as HDR images) from real-life scenes with a sensitivity and range of light far beyond what normal cameras are capable of capturing. These images can then be used for various purposes: to "light" computer-generated objects so they blend more realistically with the real world, to measure indoor color and illumination, or to be manipulated into "super photographs" that look more like what a human would see in real life than what a standard camera is able to produce. This last purpose is a very active area of research.

Another benefit of HDR images is that it is possible to modify their exposure in an interactive manner after the image is acquired. In contrast, standard cameras force the photographer to choose an exposure level before the image is taken and, if it turns out too light or too dark, the photographer must either retake the image or attempt to fix the image in ways that may degrade its quality. In addition, the photographer may be faced with the problem of a scene containing both dark elements and bright highlights. In this case, the photographer may be forced to choose between several poor solutions (for instance, taking someone's photograph against a sunset).

The most common method for producing HDR images is to capture a single scene over a range of exposure levels (say, 3-15) and then mathematically combine this collection of low dynamic range (LDR) images into a single HDR image. Each channel in an HDR image is represented by a floating-point value with high precision and exponential range instead of the typical 8-12 fixed-point bits available in most LDR images (including "camera raw"). This allows the vast range of real-life lighting to be captured in a meaningful way. In contrast, standard photographs may make it seem like the sun, a lightbulb, and a candle all have the same apparent brightness, depending on the scene and exposure levels.

Clipping is essentially when a pixel value gets to pure black or pure white and no longer has any place to go. I mean, how much more black can black get? The answer is none. None more black.

What we do is, if we need that extra push over the cliff, you know what we do? Put it up to eleven... Exactly. One louder.

Relationship between HDRI and Spinal Tap shamelessly exploited from T.S. Perry.

About AHDRIA

AHDRIA is software developed for Automatic High Dynamic Range Image Acquisition. Its purpose is to provide a professional-quality HDR imaging solution using common technology: a desktop/laptop computer and a consumer-level digital camera. The software itself is compatible with many Canon™ cameras and operates under the Windows™ operating system. AHDRIA is intended to solve or alleviate the following problems with manual HDRI acquisition: AHDRIA is responsible for image capture (i.e., controlling the camera during the multiple-exposure process) while AHDRIC is responsible for recreating an HDR image from the resulting image sequence. However, for different applications, different combinations of these tools may be used. For instance, if an image sequence is available but it was not captured by AHDRIA, AHDRIC comes with a tool, HDRI Helper, which can pull exposure information from the JPG sequence in order to allow AHDRIC-based reconstruction.



Supported Cameras

AHDRIA was written with v7.3 of the Canon SDK. The following cameras are among those that may be supported (green=cameras for which AHDRIA has been confirmed to work, red=cameras for which AHDRIA may not work, *=cameras which may only work with version 7.1 of the Canon SDK drivers): Please send me a note if you have tested the software with any non-highlighted camera. Sorry, SLR's (EOS cameras) are currently not supported.

Updates

Download

The following files are required to run AHDRIA (see Readme.txt in the first ZIP for installation instructions): Other Utilities: Older Versions: AHDRIA is available for personal, non-commercial use only. Academic use is permitted with appropriate credit/citation: "S.M. O'Malley, A Simple, Effective System for Automated Capture of High Dynamic Range Images. In: Proceedings of the IEEE International Conference on Computer Vision Systems (ICVS), New York, NY, January 2006." (A copy of this paper is available here.)

Future Work

Ideas that may be added to future versions of AHDRIA: Screenshots / Workflow

AHDRIA Main Menu
The AHDRIA Main Menu
(Viewfinder updates in real time.)

After acquisition with a minimum exposure of 1/1000th of a second and maximum exposure of 2 seconds, 12 JPG images were acquired. AHDRIA also produced the following text file: 
AHDRIA Information File
-----------------------
Shots acquired: 12
01: Exposure length: 0.00100   Relative Stops: 0.00
02: Exposure length: 0.00200   Relative Stops: 1.00
03: Exposure length: 0.00400   Relative Stops: 1.00
04: Exposure length: 0.00800   Relative Stops: 1.00
05: Exposure length: 0.01667   Relative Stops: 1.06
06: Exposure length: 0.03333   Relative Stops: 1.00
07: Exposure length: 0.06667   Relative Stops: 1.00
08: Exposure length: 0.12500   Relative Stops: 0.91
09: Exposure length: 0.25000   Relative Stops: 1.00
10: Exposure length: 0.50000   Relative Stops: 1.00
11: Exposure length: 1.00000   Relative Stops: 1.00
12: Exposure length: 2.00000   Relative Stops: 1.00
Total shutter time: 4.0067 sec.
Total acquisition time: 192.3590 sec.
Time efficiency: 2.08%
"Time efficiency" refers to the time taken to gather the images versus the time the shutter was actually open. The images were gathered at maximum resolution and quality, so most of the excess time in this case was taken transferring the images to the computer over the USB connection. The HDR image can now be constructed in two ways. The recommended method is using AHDRIC--companion software that uses the above text file to generate the HDR image from the JPG sequence automatically. The other method is to copy the "Relative Stops" information into HDR Shop (or similar software) as follows:

HDR Shop Construction
Constructing an HDR image from an AHDRIA-captured sequence manually in HDR Shop

Automatic HDRI Creation

AHDRIC is software developed for Automatic High Dynamic Range Image Creation. AHDRIC accepts the files generated by AHDRIA to produce an HDR image completely automatically--no human intervention required. It uses AHDRIA's ExposureInfo.txt as input to the HDRI synthesis process. The file it generates is in the "Radiance HDR" format, supported by most packages utilizing high dynamic range imaging.

Download

Future Work

Screenshot

AHDRIC Output
AHDRIC does not operate from a command-line interface; the only input it requires (the AHDRIA-generated text file) is retrieved through a standard Windows dialog. In this case, it took ~76 seconds to process a large (2048x1536) HDR image synthesized from a series of 12 JPGs. This would normally consume over 100mb of memory, but AHDRIC conserves memory by operating on each color channel individually.

High Dynamic Range Images

Some of these were gathered manually; most were gathered using AHDRIA. The JPG thumbnails shown with each image were produced either by viewing the HDR image at a fixed exposure, or by using Reinhard's tone mapping plugin. Use HDR View, Photomatix Basic, or HDR Shop to appreciate the linked HDR files in their full glory. (Mac users, try Photosphere.)

All of these images are ©2004-2006, Sean O'Malley. They may not be reproduced in any form or manner without permission. However, their use for academic publications is encouraged.

English Bay, Vancouver, Canada Site of UFFC'06.
Window Monkey A demonstration of high-contrast indoor-to-outdoor photography. Also, a monkey.
UH SERCC Behind the new University of Houston Science and Engineering Research and Classroom Complex. All this needs is a careful cropping and it'll make a good UFO image.
Palm Springs Site of MICCAI 2005.
Bottles: Perrier vs. Saratoga. Taken in the Grand Hyatt Hotel in Washington, D.C., during CVPR 2004 (a computer vision conference). Interesting caustics are visible in the bottles when viewed at a lower exposure level.

This image is also included in an nVidia Demo.
UH Library Art I (+Fog) Night shot of M.D. Anderson Library at the University of Houston, showing a newly-installed sculpture which projects phrases in multiple languages onto the walls of the building.
UH Library Art II (+Fog) Close-up of the above.
UH Cougar I (+Fog) Bronze cougar statue. At night. In the fog. The depth and detail in the trees is interesting.
UH Cougar II (+Fog) Bronze cougar statue, detail, with bluer color balance.
UH Tree (+Fog) It's a tree. In the fog. With rays of light... Wish I'd had a laser on me.
UH Cullen Night shot of the main administrative building at the University of Houston; taken from across a pond area.
UH Engineering Night shot of the main building of the Engineering Department at the University of Houston.
UH Architecture Night shot of the Architecture building at the University of Houston. This building, inexplicably, has a miniature Greek temple on the roof.
UH Art A couple of the art installations on the University of Houston campus. The upper-left insert shows a detail of the sculpture on the left. The lower-right inset shows how simple it is to find the location and color of all the light sources in the scene just by turning down the exposure.
Mountain Dew The green, the luminescent, the official drink of computer programmers! The inset shows some of the detail present in the scene that is almost impossible to visualize using standard photographic techniques. Not only is the filament of the lightbulb visible, but the writing on it can be read as well. This image is challenging to range compression algorithms due to its high-contrast details.
UH VisLab Part of the visualization lab at the University of Houston Texas Learning & Computation Center, where I work. Great lens flares, bright highlights, and dark shadows.

User Gallery

HDR images gathered by others using AHDRIA and/or AHDRIC and made available for viewing here (thanks!). Use HDR View, Photomatix Basic, or HDR Shop to appreciate the linked HDR files in their full glory. (Mac users, try Photosphere.)

These images are copyright © their respective authors; please contact them before reproducing these images in any way.

John Niland Scientia Building on the upper campus of the University of New South Wales, Sydney, Australia. Courtesy of Jeremy Pronk.
Kitsilano Beach, Vancouver, Canada Webcam The first HDR webcam that I know of; a very interesting application. Includes timelapse capability. Runs on a Canon A520, AHDRIA, and AHDRIC.

10-Oct-2006: This site appears to be down.
Still Life With Caustics Courtesy of Dan Reetz.
Bug's Life Courtesy of Dan Reetz. Originally captured at 7mp resolution.
Other HDRI Collections
Light Probes Panoramas Do-It-Yourself HDRI
Using HDR Images
The contents of this webpage and all files contained herein, unless otherwise noted, are copyright ©2003-2004, Sean O'Malley.