The art of HDR photography part 2
Chapter 2: HDR Challenges
See also Chapter 1: Dynamic Range & HDR
|The Ghost of Grafton|
To better understand the "Essential HDR Workflow" we need to understand some of the challenges we face creating HDR photos.
Since we're taking multiple photos, we don't only have moving elements in a single shot to deal with: any elements that move from between shots will appear "ghosted". The image below shows such a situation.
|Person moved while shooting all frames|
Note that in this case the ghosting was intentional. Ghosting is not always a bad thing; it always depends on your own goals with your images. That said, generally speaking we try to avoid ghosting as much as we can.
Chromatic Aberrations (CA) and Fringing
Chromatic Aberrations (CA)
There are several different kinds of Chromatic Aberrations, but here we dealing mainly with Lateral Chromatic Aberration, as it can be reduced using raw development software. Lets first show an example:
|CA shows most often as Red/Cyan fringes but it also can be Blue/Yellow. CA is a lens property and all lenses have this defect to some degree. The further you get from the center of the frame the more the different light frequencies focus differently on the image plane. The lens manufacturers try to correct CA and different lenses show a different amount of CA (Wide angle lenses and zoom normally show stronger CA).|
|Nearly all raw converters can reduce CA to some extent. Many years ago the French company DxO started to correct CA (and other lens deficiencies) automatically, based on the lens and camera information. Also the latest versions of Lightroom 3.x and Camera Raw 6.x (ACR) allow automatic CA corrections as part of their lens correction feature (uses camera/lens profiles). Here is how Lightroom 3.3 showed the corrected image:|
So why do we care so much about Chromatic Aberrations?
- The latest high resolution digital cameras resolve more and more of the CA that our lenses produce.
- But more importantly CA gets amplified when we merge images to HDR, multiplying all the CA of the different exposures.
|There are also cases where we see 'purple fringing' that may be related to sensor leaking between the pixels. It shows at high contrast edges. Lightroom and ACR have a "Defringe" option that also helps to reduce fringing a bit.|
HDR Challenge: Motion
Since we're taking multiple photographs movement is even more of a challenge than it is with single shots. Basically we deal with two types of motion:
- Camera motion
- Motion in the scene
For each single shot you need to take to avoid camera shake. On top of that you also want to make sure that there is no motion between the shots.
a) Using a Tripod
- The tripod, head and plate need to be sturdy
- Use a remote to avoid introducing shake to the camera
- Use mirror lockup or live view to minimize mirror slap
Note: Some photographers assume that there is no camera movement between shots if they use a good tripod. You will be surprised how much the tripod gets rattled by most camera mirrors and shutters. Even very minor movements can degrade the final HDR quality (if the images do not get perfectly aligned, more on which later).
Exercise: Try to use a 200mm lens from your tripod and experience the misalignment you get.
It's worth noting that we personally shoot a lot of HDR images handheld.
The key thing here is to use fast shutter speeds and utilize camera/lens image stabilization. The most critical exposure is the most overexposed frame because it has to have the slowest shutter speed.
The best approach is to use the highest continuous frame rate your camera can deliver. The ideal burst rates for the purposes of handheld HDR are 7-10 fps, but we also use slower cameras down to 3fps (in this case the demands for steady alignment are higher though). High frame rates ensure that the camera moves less between shots, and will reduce visible movement (ghosting) within the scene.
It should be clear by now that the key to the best possible results lies in the perfect alignment of all the images shot. In the Essential HDR Workflow chapter we'll share our personal technique to archive this goal. Here is an example what software alignment can do.
The following image shows how bad the images were aligned as shot (the image shows the different layers showing through).
OK, this was shot handheld. Ah, we hear people saying; we told you use a tripod. But this was shot on a Pacific cliff in a very strong wind. In this case shots using a tripod could well have be even worse, and would not have allowed us to shoot at the angle we wanted. In the worst case both camera and tripod could have been blown into the ocean! Here is what Photoshop could do with it:
|Aligned in Photoshop|
As this example clearly shows, software image alignment is crucial for most HDR work, especially given how difficult it is to get 'perfect' shots every time. We'll get into more practical detail in our next chapter.
At first glance you might think that aligning images using software is a relatively simple task: you just shift the image a bit until it lines up - and in fact this is some implementations actually work. But, naturally, it's never that simple. The camera can move in many ways:
- Front/back (means each shot shows a slightly different magnification)
- Any combination of the above
- On top there are also lens distortions which make it even more complex
Remember all misalignment (whether major or not) degrades your final image resolution; all your expensive Mega Pixels are easily wasted.
Here is an example of poor alignment:
I should mention that thus far we haven't usually named or recommended specific software packages. For one thing this is because the various options all improve over time, for another it very much depends on the images used (there is no universal solution). In a later chapter we'll show a method that helps us to get consistently good results.
Often you realize very late in your workflow process that the alignment was less than optimal (the above sample is of course easy to spot) It may be very subtle but we aim always for best possible alignment.
|Alignment in Photoshop CS5|
Motion in the Scene
Compared to misalignment issues, motion within the scene is far harder to correct (using processes sometime referred to as 'de-ghosting'). You will be surprised how many things move. Motion in the scene means mostly movement between the bracketed shots, and the faster you shoot (i.e. the shorter the time between shots), the fewer elements in the scene can move. Just a few of the potential 'movers' in scenes are listed below:
- People and larger animals
- Cars and other transportation devices
- Birds, Butterflies
Lets have a closer look at some of these elements.
People and animals
The good news is that these are often easy to spot, allowing you to take them into account at the point you're actually taking your photos. The bad news is that ghosting created by moving people (and larger animals) are also tough to remove, so the best solution is to analyze the scene carefully while you shoot and avoid them completely if at all possible. Some ghosting in the background (distance) is often perfectly acceptable.
Cars and other transportation devices
Being large, these are equally problematic, but typically are moving a lot faster, meaning you should be able to keep them out of at least some of the frames. Don't try HDR for fast sport :-).
Foliage is extremely tricky because:
- It's easy to forget about and hard to combat
- It can result in nasty HDR artifacts
This means that you have to watch any foliage close to the lens and in focus, as it may cause trouble.
Birds or big Butterflies
Ghosting of birds or butterflies that fly through the scene look often like sensor dust, so we treat them just like dust (i.e. we remove them using Photoshop's healing brush). Of course if the bird(s) are important to your image you may try some of the de-ghosting options in recent HDR software.
Flags are so common that they are deserve their own category. We recently tried the de-ghosting option in Photoshop CS5 and the result was quite impressive:
|Without de-ghosting||With de-ghosting|
We don't consider most clouds to be a problem because their ghosting looks just as organic as the original clouds. Of course extremely fast moving clouds can cause some problems. We recently saw some purple artifacts in the clouds using Merge to HDR in CS5. Enabling the de-ghosting feature took care of it.
Slow moving or still water (such as creeks and lakes) rarely causes any problems, but we haven't had much luck with strong ocean surf. Trying one of the de-ghosting options may improve the images.
HDR Challenge: Chromatic Aberrations (CA)
In the introduction of this chapter we explained that CA is a major problem for HDR. The best approach is to remove as much CA as possible in your raw converter. Although some HDR tools can remove CA in the Merge to HDR step, and it helps, we personally try to remove it before merging to HDR.
HDR Challenge: Camera Noise
As mentioned in the last chapter, camera noise is a limiting factor for the dynamic range of our cameras, and the reason to use exposure bracketing is to overcome this limitation. But once we start producing HDR images, we tend to open up the shadows (during tone-mapping) more than we ever did with standard processing. This again reveals the noise. There are, of course, several ways we can minimize the impact of noise on our HDR images:
- Use as low an ISO setting as possible
- Use raw converters that perform good noise removal (e.g. Lightroom 3.x/ACR or DxO).
- Use noise removal filters on the final tone-mapped image.
HDR Challenge: Source Image Formats (Raw or JPEG)
JPEG. In our personal work we hardly ever use JPEG images. Why?
- JPEGs show compression artifacts
- They are limited to 8-bit color depth, while raw images can have 12-14 bit information. More color depth allows finer color gradations
- Inaccurate white balance can degrade JPEG image quality
- Less headroom for highlight recovery
- Lower dynamic range
If you're using JPEGs for HDR the last point isn't really an issue since you can capture higher dynamic range via multiple shots.
If you don't want to deal with raw images, creating HDR images from JPEGs is still a valid technique; you just need to understand the limitations.
Raw. With the Raw images you have access to all the information the camera captured. Actually HDR files and raw images are both linear file formats. To be able to view these images both require the application tone curves.
Here is a practical dilemma. In the ideal world all "Merge to HDR" operations would start with the linear HDR data. Otherwise the HDR tools have to undo the tone curve operations performed by the raw Converters. This is why quite a few HDR tools offer the option to create HDR directly from raw (mostly using the public domain raw converter dcraw).
So why do we prefer to use an external Raw Converter such as Lightroom 3? Simply because we think that in the end these Raw converters allow better control over noise removal, CA removal (e.g. Lens profiles in Lightroom) and lens distortion corrections and offer better de-bayering and custom color profiles (e.g. DNG color profiles via ColorChecker Passport from X-Rite).
We always convert to 16 bit TIFF files before the "Merge to HDR" step, and the loss of quality due to the tone curve reversal seems quite acceptable in practice.
HDR Challenge: Lens Flare
Lens flare degrades image quality in certain shooting situations (such as sunsets). This in itself is nothing new, but with HDR we tend more often to include bright light sources into our compositions, which means flare is more prevalent.
You need to understand these HDR challenges to get the best possible quality. This list is almost certainly incomplete, but hopefully we've covered the main hurdles you'll face. Feedback by our readers is very much welcome.
This is an edited version of the first chapter of an ongoing work by Uwe Steinmueller of Digital Outback Photo, featuring his personal experiences of HDR photography, and will eventually form the basis of a book on the art of HDR photography. If you'd like to find out more about digital imaging workflow from a fine art photographer's perspective then check out the Digital Outback Photo E-book, 'The Digital Photography Workflow Handbook (2010)', by Uwe Steinmueller and Juergen Gulbins, which covers the complete digital photography workflow from input to output. The 540 page prize-winning handbook covers everything from Import to Print (and even backup) and also features Photoshop and Lightroom techniques, HDR, color management and raw editing.
© 2011, www.dpreview.com & Uwe Steinmueller.