I figured that other people who use HDR techniques on this forum will find this usefull,

This is not a fully developed test image from Death Valley, I left the horrid sensor hair in the photo for this reason!



Kathryn and I are about 3 days away from being in Utah, and I wanted to do a lot of high dynamic range pano stitching when we get there. Given the number exposures necessary for both good dynamic range and stitching, I thought I would try a little experiment, checking if I could skip exposures in the bracketed sequence and still get a high quality image. So I used this image as a test, seeing the developed results of leaving exposures out of the high dynamic range blended negative used for development.

And the results seem to show that I will need the entire exposure bracketing series to get rid of all the noise. The shutter times for the seven exposures were, 1/45, 1/20, 1/10, 1/6, 0.3, 0.7, and 1.5 sec.

Here is the results (sorry, hope you have a broadband connection!).

Hey Tim,

Cool info here. I am just starting to work with HDR, so using 6 to 7 exposures look best. Do you recommend bracketing like 1/2 stop steps, or doing full 1 stop steps?

Thanks for the effort in putting this together.

Jim

Cool info here. I am just starting to work with HDR, so using 6 to 7 exposures look best. Do you recommend bracketing like 1/2 stop steps, or doing full 1 stop steps?

The number of exposures necessary will depend on the dynamic range of what you are tring to capture. I typically shoot between 4 and 7 exposures. Exposure 7 of this series had 50% of the image fully overexposed to white.

I think I mostly bracket 1 stop between exposures. I have the Canon 5D set to 1/2 level increments, and then I move two clicks on the shutter dial between exposures going well beyond the +2 stops overexposure limit of the exposure meter.

From the test above, going 2 stops between exposures (exposure 1+3+5+7) introduces visable noise, and going 3 stops between exposures (exposure 1+4+7) yields poor results. So keeping with 1 stop between exposures seems ideal.

I should also note that the vertical strip image was slightly over sharpened to make detecting the noise easier.

Thanks for the effort in putting this together.

No problem.

This is pretty interesting info.

One thing I will tell you, though, is that what you are proposing to do will create quite a bit of post-processing work, and fill a lot of CF cards. If you have a professional stitching program, it should be manageable, however.

One other thing I would watch out for is how quickly you can do this in fast-changing light such as sunsets or sunrises, and also in fast changing conditions like wind-blown clouds. Having to make 7 different exposures for each "shot" in your pano will slow you down, and you'll want to be careful not to take too long so that elements of the image don't change in between "shots".

Make sure to post your results when done. Can't wait to see 'em.

Kory,

Thanks for the tip on movement and time problems,

I am planning to do only 3-4 horizontal shots per pano during the critical moments, and when the light is more relaxed probably 5-8 vertical shots per pano. Funny thing is that really it's my wife that is more concerned with the time than I (because she won't have the good camera). I also just got 2 extra 4 GB flash cards, so I should be good on storage for a morning or evening shoot with a total room for about 650 exposures.

If all works out, I will definatly post the results...

Hi Tim,
I'm just 2 days away from leaving Utah (for the Tetons), but am interested in your workflow. I'm not a fan of HDR yet, but like to keep up on it. If you aquire, say, 5 bracketed shots each of a 4 shot pano, for a total of 20 images, do you make 5 seperate 4-shot panos, 1 for each shutter speed, and then blend them all into an HDR later, or do you blend each of the 5 versions of each seperate shot into their own seperate HDR, and then stitch them all into a pano at the end? How do you assure that your control points will work out exactly the same for each of the 5 panos that are to be blended into the 1 final HDR pano, epsecially if things like clouds, sun, water, or branches may have moved between shots? Sounds quite difficult to get it to all line up perfectly.
-David

If you aquire, say, 5 bracketed shots each of a 4 shot pano, for a total of 20 images, do you make 5 seperate 4-shot panos, 1 for each shutter speed, and then blend them all into an HDR later, or do you blend each of the 5 versions of each seperate shot into their own seperate HDR, and then stitch them all into a pano at the end? How do you assure that your control points will work out exactly the same for each of the 5 panos that are to be blended into the 1 final HDR pano, epsecially if things like clouds, sun, water, or branches may have moved between shots?

I avoid all these problems. Other than the stitching step, my workflow is exactly the same as it is for doing a non-stitched shot. Of course getting all the 20 shots before things move too much is always a challenge. But this process only takes about 4 times as long as my current non-stitched process, so that's not much of a problem. My exposure blending can deal with the small amounts of movement I see without any major problems.

When shooting, I take the full bracket of exposures for each shot before moving to the next shot.

Here is my workflow for a 4 shot pano with 5 exposures per shot,

1.) Convert the RAW files to DNG files.

2.) Load all the DNG files at once into ACR and set all files to the same White Point and Chromatic Aberration setting, and zero out all other settings (to get an-undeveloped sequence of files).

3.) Run the batch version of FFDD bracket. This blends the 5 un-developed exposures for each of the 4 shots into a high dynamic range blended digital negative.

So now I have 4 shots which have a large amount of dynamic range saved in 4 16-bit tiff files. Note that unlike doing a Photomatix or Adobe HDR with the local contrast adaption option, my blended negatives will not have any artifacts or distortions, they are infact as close as can be to what a single high dynamic range negative would look like if such a camera existed.

4.) Stitch these 4 high dynamic range shots together into a stitched negative.

5.) Create a copy of the large stitched negative, and shrink it down to 1200 pixels wide.

6.) Develop the 1200 pixel wide copy in Photoshop using curves adjustment layers and graduated layer masks to adjust the brightness and contrast. This allows for very fast development. Tring to develop the full size image would be very slow.

7.) Resize the 1200 pixel wide copy to the size of the large stitched file. Then paste in the large stitched negative below all the adjustment layers.

That's it. No HDR worries!

Hi,

I tried HDR with bridge and CS2 and the images looked terrible.. the sky was horribly over exposed in the final image.
Could you guess what am I doing wrong? If you want to get the whole dynamic range you need to take exposures for the foreground where the sky gets blown ..( i really don't understand how photoshop deals with those blown out areas)

Many thanks

Jane

I tried HDR with bridge and CS2 and the images looked terrible.. the sky was horribly over exposed in the final image.
Could you guess what am I doing wrong? If you want to get the whole dynamic range you need to take exposures for the foreground where the sky gets blown ..( i really don't understand how photoshop deals with those blown out areas)


Jane,

You are right in that if you want to get the full dynamic range, you do need to take exposures where the sky is fully overexposed to white. When CS2 blends, it uses the areas that you exposed for the shadows to create the HDR image.

You will also have to have at least one exposure that has no amount of overexposure (meaning that the sky is properly exposed, or even slightly underexposed).

Also the settings used in Adobe Camera RAW for developing the exposures before creating the HDR image can also have a large effect on the output. Some people develop the raw files and then blend, while I blend then develop after (meaning my raw images have brightness/contrast/etc set to zero before blending).

If you have the exposure range right, then the problem might be in the 32bit to 16bit HDR Conversion step. Setting the Exposure too high in the "Exposure and Gamma" method will clip your highlights causing overexposure. The other 3 methods including "Local Adaptation" tend to add artifacts (graying effect, halos, glow, etc).

HDR sure gets complicated doesn't it.

Hope that helps a little.

Tim, could you share some more insights into the FFDD method you mention opposed to photomax or the CS2 HDR option. Benifits, weaknesses and so forth

Very impressive results so far, far better than what Ive ever been able to achieve with CS2, though some of that of course is due to my inexperience, but if theres a better method, Id love to hear about it

Brutus,

Photomatix and CS2 HDR are more geared towards finding a way of reducing the dynamic range of an HDR image. In essance compressing the dynamic range to something that can be output on paper. Both Photomatix and CS2 HDR use some form of local contrast adaption to do this. BTW, the CS2 Image->Adjustments->Shadow/Highlight tool can be used to produce a similar effect on a single exposure (click the Show More Options box to adjust the radii). It is this local contrast adaption which produces the HDR halo/glow artifact.

FFDD is more geared towards reducing the noise in a single proper exposure, and then providing darkroom like tools to develop this exposure in Photoshop. FFDD can be concidered an HDR tool because in reducing the noise in the dark areas of a single exposure (that has no highlight clipping), it is extending the dynamic range of the shot.

With this extended dynamic range in the darks of an FFDD blended negative you can do things that just are not possible in developing a single exposure. Here is a 100% of a dark region of a blended negative (might need to up the monitor brightness to even see the detail in this shot),



Here is the same shot with a levels adjustment (clipping highlights) to show only the darkest stop of from this same crop of a 16-bit blended digital negative,



The traditional and natural method of dealing with a HDR scene is to use a GND (graduated neutral density filter). With the huge adjustment latitude provided by a blended negative, I skip doing the GND in the field, and just do it in the post processing.

Hi,

It was the raw files set on the auto settings that was the problem.

Thanks


Jane