 |
|
 |
 |
 |
 |
  | We would like to import our HDR composite into an image editing program, but there are few that recognize pixels with floating point values. I have experimented with two formats: a 16-bit gamma-compressed tif format, and the OpenEXR format, developed for special-effects in motion pictures.
The pixel values in the composite Vega image range up to 5*10^8, which exceeds any reasonable compression to 16-bits. The OpenEXR format uses a "half-floating point" numeric type making a smaller file than a regular ("single-precision") floating point value would. The developers of this format (Industrial Light and Magic) have made the specification public and it is now supported by a number of image editors including Photoshop.
The composite image was saved in the OpenEXR format after normalizing such that a full scale digital count (65535) in the base exposure image (30 seconds) was scaled to 1.0. When the image opens in Photoshop, this is what appears:

|
 |
 |
 |
 |
 |
  | Photoshop imports the EXR format into a 32-bit per channel image, and displays an 8-bit window out of the 32 bits. To prove that this is indeed a high dynamic range image, select the "32-bit Preview Options" from the View menu. It is not clear how Photoshop color manages its rendering of the data, but my guess is that it applies a gamma=0.5 transfer curve to approximate a generic viewing device (light intensity = data value ^ 2). The exposure values appear to be in f-stops, and the gamma slider is relative to the initial (gamma=0.5) built-in value. Here are some viewing variations.

|
 Should be the equivalent of the 30 second (linear tif) exposure.
|

|
 The equivalent of the one-second exposure, but at a gamma of 2.0 encoding.
|

|
 Decreasing the gamma reveals more of the low level signal.
|

|
 An attempt to compromise between avoiding saturation and bringing low level detail.
|
|
 |
 |
 |
 |
 |
  | The controls in Photoshop for this image format are still rather simple. To perform any editorial work we must convert to a 16-bit or an 8-bit format, in which all of the customary tools may be applied. Here is one rendition. The diffraction spikes are colored is an artifact of the sensor's color sampling mosaic. The loops are artifacts from the imperfect deblurring and alignment. The 9th and 10th magnitude stars are visible in the same view as Vega, a 0-magnitude beacon. This scene shows a dynamic range of objects whose brightness differs by 100,000:1 (!)

|
 |
 |
|


 |
 |
 |