I have been experimenting with Zero Noise, a really useful app written by forum member Guillermo Luijk ("_GUI_" avatar). You can learn all about the app and download it here:
Some of Guillermo's articles are translated from Spanish into English, others I usually just paste the URL into Google Translate. Works well enough.
Anyway, for the uninitiated, Zero Noise takes a sequence of RAW image exposures and combines the best of each exposure into a single 16 bpc TIFF. Unlike most HDR apps that combine exposures into a single HDR dataset by averaging pixels across exposures, Zero Noise segments the exposures and blends the best pixels from each exposure into a single, 16 bit dataset. This is similar to Enfuse or Tufuse, but with the benefit of greater bit depth. Also, the image remains sharp and the color remains consistent.
The absolute best part is that the shadow data where noise typically rears its ugly head when pushing data around in post becomes, for all practical purposes, noise free. With 16 bits of data and no noise in the shadows, you can tonemap the image with reckless abandon and not destroy the histogram. The data will be combined in a way that the resulting TIFF will appear as the darkest exposure in your sequence. You will have to perform some post processing to bring the image out of the data, but, no worries with no noise in the shadows you are good to go.
The idea is elegantly simple, the app is pretty straightforward, but it helps to review the workflow for those who have not tried this app.
Shoot your image sequence
Here is a composite of three exposures I shot today of a high dynamic range scene - bright afternoon sun bathing the field and trees in the background, with gothic arched stonework creating midtones and deep shadows, especially up in the dark brown roof of each arch. Incidentally, this was taken at Washington's Chapel, in Valley Forge National Historical Park. This is the kind of scene that would be a candidate for HDR, with the exposure sequence combined and then tonemapped in whatever app you choose. In fact, I have done this previously, with less than natural results:
Anyway, I shot these three exposures and combined them into a single 16 bit TIFF using Zero Noise.
Zero Noise - The Basic Workflow
So here is Zero Noise in Linux, running on my Mac Pro under VMWare Fusion:
Let's orient ourselves according to the numbered areas listed in the figure above.
1) Image Window - this is where a preview of the RAW image that you feed into ZN will appear. You can interact with the image to pick the white balance area, for example, using the "Patch" option in the Raw development tab (5).
2) ... - this button brings up the file browser - select all of the RAW images you want to feed into ZN and have combined.
3) Status Window - gives you feedback on what is happening.
4) Raw Files Selected Window - gives you information about the processing and the results of that processing.
5) Raw development tab - this is where you start your workflow. Here you can select the white balance method (I chose Daylight preset). If you choose Patch, you can click and drag a rectangular or circular area on the image as a neutral reference. If you know the RGB WB coefficients, you can enter them manually. This WB data will be fed into the RAW converter that ZN uses to "develop" each RAW image prior to combining them. ZN uses dcraw to convert the RAW images to 16 bit TIFFs.
The DCRAW command line that will be fed to DCRAW is shown below (area 8 ). Briefly, the options shown in this example are automatically generated by ZN and include:
-v (verbose output)
-4 (16 bit output)
-T (TIFF output)
-q 3 (high quality output)
-o 1 (sRGB output color profile)
-r N1 N2 N3 N4 (white balance coefficients)
You can enter your own additional command line flags here. If you enable the "HL" button (highlight recovery) you will see an additional flag "-H N" where N will be 2, typically.
Leave the Saturation Level and Demosaicing Algorithm at their defaults, but feel free to experiment. The Saturation Level is the level at which the sensor saturates and varies by camera. DCRAW knows the typical values for this and will use the appropriate value automatically.
BOTTOM LINE - after you have loaded your images, you can pretty much use the defaults (maybe with the exception of using a preset WB where appropriate, or the Camera WB "as shot" equivalent). The Develop Button, grayed out here, will be active - hit develop and let the develop process take place. ZN will develop the RAWs into TIFFs, and will write these TIFFs to the same directory as the RAWs.
Then switch to the next tab - Calc EV/Build Map - area 6 in the figure.
6) Calc EV / Build Map
Here is where the developed RAWs (now 16 bit TIFFs) will be thresholded and segmented. THe result will be a grayscale map of the image areas that will be extracted from each image and used for the final composite image.
a) click on the Calc EV button. This will make ZN figure out the relative exposure of each RAW (in EV). You will see feedback in the message windows (3) and (4).
b) click on the Build Map button. This will build the grayscale segmentation map, a PNG file that will be written to the same directory as the RAWs, etc. This map will be used in the final combination step in the Image BLending Tab.
Just a note - the sliders here control how the map is built - the threshold, anti-ghosting and progressive blending sliders will vary how much data is taken from each source image. Play with them. Early on, just accept the defaults.
Here is an example of the Progressive Blend map for the above composite:
The number of gray scale values will be equal to the number of images you feed into ZN. Black represents the pixels taken from your lightest exposure, white represents the pixels taken from your darkest exposure.
7) Image Blending tab
Here is where the magic happens. For now, just hit the "Colour" button to combine the data into a single 16 bit color TIFF. The blending map that created previously will be automatically entered in the field at the top of the window. As you can see there are other options which, frankly, I have not used. This operation will write a TIFF file to your drive that will have a filename that reflects a serial number and some info about the color space and gamma used in the production of the image. Recall that the image ZN produces will be equal in luminance to your darkest exposure in the sequence you feed into ZN.
Here is what the output image of the above exposures looks like straight out of ZN, opened in PSCS4. Note, when you open the image, you will be prompted to assign a color profile - make sure to assign the same profile you specified in (5).
Pretty dark, huh? Never fear.
Here is what would happen if you tried to tonemap the lowest exposure RAW into a decent image, in terms of luminance distribution:
Here is what you can get with a bunch of massaging of the ZN TIFF:
Taaadaaa! Much more natural than your typical tonemapped HDR, in my opinion - for whatever that's worth!
This is a brief overview of Zero Noise. Next, I will put together a little tutorial about how to tonemap the resulting TIFF. For starters, check this out, from Guillermo's site, translated through Google into English:
Have fun and experiment!
PS - Guillermo - I apologize if I have screwed the pooch on anything here. Feel free to critique and clarify!