Posted 17 November 2006
Reader Commentaries and Responses on "High Dynamic Range Photography" by Harrison Eiteljorg, II
This exchange between Irwin Scollar and Harrison Eiteljorg, II, followed the publication of the article on High Dynamic Range photography ("High Dynamic Range Photography" by H. Eiteljorg, II, CSA Newsletter, XIX, 2; Fall, 2006: http://www.csanet.org/newsletter/fall06/nlf0603.html).
Comments and responses are listed in the order that they were received, with the most recent last. The latest addition was received 29 Sept 2006.
From Irwin Scollar (23/09/06):
Since you saved your images in JPEG, you probably reduced their dynamic range to 256 levels, thus removing high dynamic range.
You might try saving either to 48 bit Tiff or since you are using a Nikon, to NEF (Raw format). Both preserve the dynamic range seen by the sensor. If you have a Canon, you can use CRW (Raw format) for saving.
I don't have Photoshop, so I don't know if it can handle these formats and preserve their dynamic range or not.
My freeware CastCorrect program can read and process all of these formats. It also does Retinex and ACE correction in addition to its normal function for removal of color cast from aged positives.
There is a vast HDR literature, and there are a number of free utilities which can handle images in various formats which you can find on the links. If you have Matlab, you can even run the code offered by Meylan which works very well indeed on the CRW imagery she used for testing, but it can't read NEF.
Harrison Eiteljorg, II, wrote back to Irwin Scollar as follows (29/09/06):
Just finished processing 5 NEF images. No better. Since the set-up was slightly different, the images were as well, but, once again, the process left out the details in the brighter areas, showing no texture for the wavy cardboard on the floor. (Interestingly, I remain unimpressed with the way my Nikon D100 changes shutter speed. Turning the dial yields unpredictable results and worries me as to image stability.)
and to various lists as follows (29/9/06):
A second experiment was conducted after Dr. Irwin Scollar emailed a note suggesting that the use of the JPEG image format might have had a negative impact on the results of this experiment. Not only was he correct, it was a procedural error I had recognized and regretted but not had time to correct. Another set of 5 images -- using as nearly the same set-up as possible was taken with the same camera and lens but storing the images in the Nikon NEF -- "raw" image -- format. This format preserves more information and should have been used in the initial experiment. Although the file format preserves more image information, including superior color representation, the results were not better. In the bright areas the H.D.R. image was washed out and lacked detail. The results are not illustrated here since that would require a complete new set of images while adding no real information. My sincere thanks to Mr. Scollar for reminding me of the need to make the added experiment. (Note added 29 September 2006.)
Irwin Scollar responded further (29/09/06):
I have a suspicion that Photoshop is converting the NEF to 24 bits before doing anything with it. In my own software, I'm using a commercial image file handler to read and write NEFs, and it may do the same thing. I have the source code for my thing, so I can see what is happening. If my suspicions are confirmed, then that would account for not seeing much happening. Another source of possible trouble is the display hardware, since this too is usually only 24 bit, as are LCD monitors as far as I know.
As an intermediate step when you have a moment, you might try downloading the trial version of the new version of Nikon's Capture NX and try that rather than using Photoshop. It's possible that they make optimum use of their own file format. I have heard that it runs for 30 days.
Editorial Response:A new experiment will be conducted for the next issue of the Newsletter, with Eiteljorg and Scollar working together on the image processing.