Lightroom Photo Merge DNG versus 32-bit Floating Point TIFF

January 19, 2017  •  Leave a Comment

UPDATED January 20, 2017: The original post from July, 2015 follows the page break below.

In July, 2015 I posted results from my early testing with Lightroom 6.1 Photo Merge to HDR feature in comparison to 32-bit floating point TIFFs created in Photomatix. Since then I have continued to use the Photo Merge tool with mixed results of success and disappointment. This post is to share an updated set of side-by-side comparisons from images I shot just yesterday so these results are definitely current for Lightroom CC 2015.8. In the context of this post, where I speak of DNG files, I am referring to DNG files that were created within Lightroom's Photo Merge module only, not DNG files converted from any other photo RAW. And where I speak of TIFF files, I am referring to 32-bit floating point TIFFs created in Photomatix.

Each of the updated images below include easy to understand call-outs with the specific comparisons so I won't re-type all the comparisons here. But what you will see are examples where Lightroom DNGs lack detail retention, creation of false aberrations / color banding, de-saturation of color, and creation of artifacts.

For the record, let me be clear that I'm not saying all is bad with Lightroom's DNG/HDR files. I use the Photo Merge module on a regular basis with nature / landscape HDR images and I usually love the results. In my original testing in July, 2015 I shared results that spanned scenarios including interior, landscape, and architectural elements. And since then I have seen improvements, particularly in areas of noise control, or at least the merge process not actually creating noise. I still experience a lot of artifacts being created from Lightroom's de-ghosting algorithms, but things have improved. Best yet, the entire Photo Merge process seems to run faster now. But unfortunately I am not seeing improvements to the results I am getting with interior HDR images. 

On of my longest-running and personal favorite projects is creating HDR images of church sanctuaries. And in the purest sense of high dynamic range scenes, many of those spaces are as 'brutal' of an environment as you will ever encounter. When I photographed this space yesterday, I did so with the window shutters open and as you see here, with the shutters closed. The window glass is tinted in different colors but is essentially clear glass. The right side of the building faces due south and this was shot on a bright day with the sun full-on the south side of the building. As bad as the results look here, the results are worse with images shot with the shutters open.

Please note the images shown here are not finished images. The DNG and the TIFF were created from the same set of 15 exposures, captured at 1.0EV stops, ranging from 1/1250-second to 13-seconds. After each file was merged and imported back into Lightroom, I performed basic toning as a 'starting point' for final processing, and that's what you see here. Both files are still rather flat but both are ready to go into Photoshop for the final steps.

Here are the new comparisons. After clicking on the first image you can use left and right navigation in the photo viewer to see all images.


Original post from July, 2015:

In a previous article titled Maximizing Your HDR Photography I explained my approach to HDR photography and my method for creating 32-bit floating point TIFFs using Photomatix (see Post Processing segment).  Since posting that article I have been asked about using the Lightroom (Lr) PhotoMerge HDR feature introduced with version 6.  At the time I had yet to test the PhotoMerge feature thoroughly, but after much excitement and anticipation I have compared the two file formats head-to-head and the results are in!  In a word, my initial impression of the Lightroom 6 "HDR DNG raw file" is this: disappointment.

Prior to LR6 being released I had read the new PhotoMerge HDR feature would generate a 32-bit floating point TIFF (which turned out to be false). I was very excited to have another alternative for creating floating point TIFFs, and having that new tool built right into Lightroom sounded awesome for further streamlining my workflow. And while the workflow in itself is simplified quite a bit using the PhotoMerge feature, I find the quality of the resulting image surprisingly disappointing in my experience so far. 

Without further ado, let's compare several real-world examples. All files for this comparison were created equally. The photography processes that went into these projects are verbatim to that described in the aforementioned article and both files were created using the same set of exposures. Once I had the 32-bit TIFF and the Adobe HDR/DNG files assembled, I developed both in Lightroom 6.1 to match them as closely as I possibly could. All of the comparisons made here are 1:1 screen shots of the two files side-by-side as displayed in Lightroom's comparison view. In all examples the Lightroom PhotoMerge HDR/DNG is displayed on the left, and the Photomatix 32-bit floating point TIFF is displayed on the right. For the purpose of this article I'll refer to these as "DNG" and "TIFF" respectively. Click on all thumbnails for a larger view.

Figure A: 18 exposures at +/- 1.0EV ranging from 1/1000 to 90 seconds.
Figure A

Figure A at left is a classic example of my interior HDR photography. This type of space can be difficult to photograph considering the vast differences between highlights and shadows, not to mention trying to manage color correctness given the typical mixed lighting. Having said that, it is paramount that your master tone-blended file be as clean as possible to ensure superior image quality in the end. Figure B below takes a close look at the pros and cons of the DNG and TIFF files for this space. To quickly address each point:​

  • Note A: It is typical to have light spill and color spill when photographing stained glass windows. It is not typical for that light spill to form with hard edges as it has in the DNG file. The TIFF file shows a more natural and smoother transition in the light and color spilling from the glass. Note this occurs around all of the windows.
  • Note B: Looking at the DNG file, notice the band of red color cast that is collected across the top of the pillar. Notice that color cast did not occur in the TIFF file. There was certainly no red light in that area. It actually appears as if the Adobe tone-blending algorithms created a chromatic aberration at that spot. 
  • Note C: Similarly to Note B, there is an orange and red halo emitting from the lamp in the DNG file. Clearly the TIFF file looks far better with no color cast and with nicer, crisper edges around the lamp's framework.
  • Note D: Look at the detail in the glass in the DNG file. Or more accurately, notice the lack of detail in the glass in the DNG file. Note all of the glass looks the same. As I mentioned earlier both files were created with the same set of exposures. And although the highlights were protected with a 1/1000-second exposure, the DNG file does not contain the tonal range needed to completely dump the highlights in the windows. Aside from all the shortcomings listed in notes A through C, blown out highlights is a showstopper for me where my interior photography is concerned. Clearly the TIFF file wins this contest.
Figure B
Figure B



Figure C: 13 handheld exposures ranging from 1/8000 to 1/40 second
Figure C

Figure C shows the final image of a 13-exposure project. While the final image was processed in black and white, the examples used to compare the DNG and TIFF files are shown in color to illustrate the original condition of each file. I am providing two examples for this image in Figure D and Figure E below due to the multiple issues I have observed with the DNG file.


Looking at Figure D:

  • Note A: One of the things I've admired in Lightroom is the program's ability to reduce color noise. But in using the PhotoMerge HDR feature I find it curious that Lightroom actually generates color noise. Look in the shadow under the eave and notice all of the green color noise, or green speckling. Note this doesn't occur in the TIFF comparison.
  • Note B: We're looking at detail and contrast in the shingles. The TIFF file wins again.
  • Note C: Notice the clarity of the glass in the TIFF file. The DNG file is rife with color noise and luminosity noise.
  • Note D: Again, lots of color noise and luminosity noise in the siding of the DNG file compared to the smooth color and tone in the TIFF. Also note the clean edges of the siding in the TIFF. 
Figure D
Figure D

Picking up with Figure E we move to the cemetery for our next set of comparisons, where unfortunately we find much of the same issues as with Figure D. The issues are:

  • Note A: Notice that orb of light on the grave stone. Why did the DNG create that orb of light? There were no lens flares in any of the exposures used for this project. I'll take the TIFF ...
  • Note B: Color noise, luminosity noise, and more color noise and luminosity noise. Photomatix clearly does a far superior job in reducing luminosity noise in any underexposed file.
  • Note C: If you look closely you can read the inscription on the grave stone in the TIFF file. The DNG completely lost this detail in the noise.
  • Note D: You may have noticed in the photo caption that I shot this handheld. Being so, both programs had to align the files when creating the composite. The DNG actually has a slight edge here for aligning the detail in the leaves. Finally, I see something positive in the DNG! 
Figure E
Figure E



Figure F: 9 exposures at +/- 1.0 EV
Figure F

Although I prepared several more examples for this comparison, as I continue with this article I realize I'm just beating a dead horse and that I'm only rehashing the same issues over and over. But in closing I would like to provide one last example that seems important to me. Figure F shown at left is a scene that was practically made for HDR photography. This image was assembled with 9 exposures at +/- 1.0 EV.  The reason I feel this image is important is that it has a characteristic that we deal with often in HDR landscape photography which is motion of objects and de-ghosting in post processing. 

Figure G below shows artifacts in the DNG file from what I suspect is Lightroom's de-ghosting algorithm. I know for sure the clouds were moving and I think it's likely there was a slight breeze in the trees. The artifacts in the leaves shown in Note A of the DNG file are inexcusable. Not only is the artifact noise horrible, the colors of the leaves are completely different from their surroundings. The same condition is found throughout all trees in this image. Finally, Note B shows a fair amount of artifact noise around the edge of the clouds. You might guess all of the cloud edges look this way. 

Figure G
Figure G

So, in closing, I think it best that I let the pictures speak for themselves. I do have high hopes that eventually Adobe will work out these kinks. But as it stands, at least based on my experiments here, it is my opinion that Lightroom isn't ready for prime time with the PhotoMerge HDR feature. I simply place too high of a premium on image quality over convenience to consider PhotoMerge as a go-to tool for my HDR photography at this time. Here's to hoping ...

Until next time, happy snappin'!


No comments posted.