Creating High Dynamic range images

Creating HDR images

This post outlines a few different methods for creating HDR images.
It continues where Part 1 and Part 2 of Image based lighting left off.

Mirrored Ball Method

One of there easiest, and least effective ways to make an hdr image, is use "mirrored ball" approach. In this case, photograpsh need to be taken from a single perspective, at 10+ different exposures and merged in Photoshop or HDRshop.

Surprisingly, when "unwrapped", this provides almost 360 degrees coverage. This approach also shows imperfections in the sphere itself. Note the scratches on the surface metal, the distortion of the surface, not to mention the fact that the photographer and his tripod are in the middle of the picture.

This is the jpg version of the resulting image in it's original "Mirrored Ball" layout.

Panoramic Transformations

Differnt 3D packages expect IBL images in different layouts. Houdini's environment light expects a LatLong layout (also called sphererical mapping). Below is the "unwrapped" version of the image.

In this case, HDRshop was used to convert the "Mirrored Ball" layout to "LatLong". Artifacts such as the occluded area facing away from the camera, and the photogrphers reflection, were corrected in Photoshop.

(HDR version Here).

Wide Angle Lens Method

Another approach is to use an ultra wide angle lens. Depending on your camera, a 4-10mm lens can provide a nearly 180 degree field of view. With this much coverage, you can capture 360 degrees from 3 perspectives, which are later merged and stitched in Photoshop.

The trick with this approach is that the 3 perspectives must each be photomerged into their respective hdr images. Then the lens distortion must be removed. Then the three images need to be stitched together. This takes a lot of cloning and image manipulation, but the resulting image is high quality. Compared to the mirrored ball anyways...

Using a Pano head

By far the most labor intensive method is to use a panoramic head. This involves taking photos with a standard lens, from 48 perspectives to capture the entire 360 field of view. Keep in mind that you need 10+ exposures of each perspective. Then you need to merge and stitch the results.

This is for the hardcore, to be sure.

People using this approach are known to hack their cameras firmware to automate the bracketing of shots. High speed and high capacity memory cards are required to handle the volume of RAW photos.

The time required to take so many photos usually results in ghosting artifacts: this is caused when objects such swaying trees and grass are in different positions from frame to frame. Lighting conditions can change with a passing cloud etc.
Usually dedicated software is required to handle the stitching.

Using a specialized camera

There is a company called Spheron which produces a specialized camera for just this task. For the ultra high end, this is probably the way to go.
All Topics
Blog posts
Press releases