Crepuscular Rays - HDR

As mentioned in previous posts, I've been experimenting with stitched HDR images. A couple of weekends ago I was returning from a no-show sunrise at Stony Creek Backwash when the clouds split open revealing a Crepuscular Ray's over the city. Given the overcast sky and the bright rays of sunlight, there was no way I could correctly expose a single image:

crespuscular rays, stonycreek backwash - metered exposure

Recognisng the HDR image opportunity, I bracketed 9 shots at one stop intervals (-4, -3, -2, -1, 0, +1, +2, +3, +4) for each of the four frames in the sticth. Ideally I would have liked to bracket at 2 stop intervals but this is not possible with the D200. In the end I discarded all the images except for the -2, 0, +2 sequence. The next challenge was to tone map the image.

Using Michael Freeman's informative "Mastering HDR Photography" as a reference, I compared PhotoShop CS3's HDR functions with PhotoMatix and FDR tools. Although I admire much of the obviously HDR tone-mapped images that appear on Flickr, I was looking for a tool that gave me the most photo-realistic results. My other criteria for selecting a tool was usability. Of the three tools I found FDRTools the most intuitive program to use. I found PhotoShop's curves local tone mapping option to difficult to use and PhotoMatix' interface too complex. As it turned out after much experimentation, FDR tools also gave me the most photo-realistic results too.

crepuscular rays, stony creek backwash - hdr version

The workflow I followed for the image looked like this:

Capture

• Four frame stitched image @f11 - bracketed -4, -3, -2, -1, 0, +1, +2, +3, +4 (36 shots in total)

Pre-processing

1. Process raw images in PS Camera Raw (White Balance as shot) - save as Tiff files.
2. Tone map each of the four frames in FDR tools - Compressor 7, Contrast 5, Smoothing 2.
3. Stitch using AutoPano Pro - save as PSD file.

Post-processing

1. Set white point using Levels.
2. Additional tone mapping with Shadows/Highlight.
3. Modify yellow and green tones with Selective Color.
4. Apply a lighter curve to foreground, masking out the sky with a Quick Selection.
5. Apply an Unsharp Mask for the midtones - 20%, 120 radius, 0 threshold.

Although not totally happy with the end result - the image still looks obviously HDR - I am pleased that I spent the time experimenting with the process. For larger stitched images (22 frames), bracketing five-nine images per frame may not be practical. But for smaller images and where the light dictates it, I'm looking forward to applying the technique in New Zealand in a fortnights time.

Large scale digital process

The following is a summary of my current large scale digital process based on the Lake Mokoan trip and West Gate Bridge series. I expect this process to continue to evolve as I shoot in new locations, gain experience and refine my technique. The next iteration of this process is likely to occur when I visit New Zealand in December.

Capture

The essence of my current process is to capture up to 22 images in a two row panorama (with 50% overlap) that yields a large scale digital image of 20” x 60” @300dpi or 14 images for a 20” x 40” image. Although some of my images are considered full panorama’s at a ratio of 3:1, I am more interested in capturing images at 2:1 ratio (similar to widescreen cinema) which I consider as large scale digital photography rather then panoramic digital photography. I will have no hesitation in cropping a full 22 image capture to the 2:1 ratio if the composition is improved.

This is all captured with my sharpest prime lens, the inexpensive Nikkor 50mm 1.8 AFD. I liken the process to digital scanning of the subject with post-processing providing the final composition. What I do ensure during capture is the thirds rule for either the foreground or background by overlapping the image 50%. For example if the sky is the feature then the bottom of the vertical frame is on the horizon for the first row, I then overlap this by 50% for the foreground. I capture in RAW and save a basic JPEG which I use like a contact sheet for selecting images for full processing. As print is the final output medium, I always shoot Adobe RGB.

Refinements of this technique including HDR capture, which I am currently experimenting with and maximising the DoF by using Helicon focus for which I have not experimented with. I could also introduce a RSS nodal slider but as yet have had few problems with parallax due to the efficiency of the Autopano Pro algorithms.

Pre-Processing – Camera Raw, Autopano Pro

I liken the Pre-Processing stage to the development phase of a film negative. After being frustrated by CS3’s ability to manage multi-layer stitching I discovered Autopano Pro. One of the key features of Autpano is the ability to automatically detect images that are part of a Panorama and stitch those images into a large scale digital image. The software has an automatic color correction feature, which can be tuned manually. Autopano can also use bracketed pictures for a HDR (high dynamic range) workflow. The program supports many file formats, including raw files from Nikon’s and Canon’s. One of the main features of a View Camera is the ability of the tilt-shift to maintain perspective. After experimenting with taking multiple shots on a grain silo, I found that the Autopano Pro software was able to correct the verticals with minimal loss of detail. As you can see from the image below Autopano is not limited to horizontal stitching tasks.

grain silo, tallygaroopna, originally uploaded by Andrew McDougall.

When I first download the images I do I quick sort by separating out the jpegs and getting Autopano do an automatic sort of the images. I always separate each image with my hand over the lens. Once Autopano has detected the panos I am able to select suitable images for post-processing and discard the rest. I then separate out the NEF’s for post processing based on this sort. In Adobe Camera Raw, I will apply global color balance corrections if required and any exposure tweaks before saving the images as Tif files. As the rendering can take some time (30-60minutes), I will often set the rendering to batch process over night. Each rendered file is saved as a PSD file at up to 500MB ready for post processing in Photoshop.

Post-Processing – PhotoShop CS3

Following along with the film analogy, the post-processing stage is like the printing phase in the darkroom. Once the photos have been rendered in Autopano as 16 bit PSD files, I open them in Photoshop. At this stage my workflow is relatively simple, I select from the following techniques to post process the image:

Global corrections:

• Fine tune the White Balance with Match Color – often unnecessary if corrected in Camera Raw, but always worth checking.
• Set the White Point using levels – not required if image includes a bright background like a sky that requires local modification.
• Adjust the contrast using Curves – as for setting the White Point.
• Use Selective Coloring to bring out the color - the RAW capture will be flat when compared to the JPEG.
• Use Hue/Saturation to increase the saturation – as for Selective Coloring
• Experiment with Filters – worth a go

Local Corrections

As mentioned above, the contrast and white-point may need to be set locally when there is a sky involved especially at sunrise and sunset when the area around the sun is clipped (as it should be). I use a number of techniques to alter the contrast including:

• Blending to images (either two shots exposed for the sky and foreground at time of capture or two shots processed at different exposures in Camera Raw – similar to pushing or pulling the negative in film development).
• Using the Magic Wand to mask out the sky or foreground (this is quick but can also be laborious with fine detail and needs to be feathered)
• The current technique I favour for complex horizons is duplicating the channel with the most contrast, choosing Image/Adjustments/Threshold and then applying the masks to either the Levels or Curves layer. The same mask can be inverted for the contrasting portion of the image. It is important to feather the selection (5px) for a smooth transition.

After working through the above techniques I save the file as a layered image so that I can always come back to it – funny how 24 hours later things look different!

Output – Print & Web

With the flattened PSD I choose Image/Shadow Highlight and check to see if this improves the image – I usually find that 10% for both the Shadows and Highlights is a great finish – though there are no hard and fast rules. I then duplicate the channel with the most contrast and follow the smart sharpening workflow to sharpen the image – I may also compare with Filter/Smart Sharpen at 100% 0.5px.

Finally I use Photoshop’s File/Script/Image Processor to create JPEG's at Quality 11 SRGB and RGB as well as SRGB quality 5 for the Web (sometimes this image may need to be tweaked further for the web). I will also create TIF file for printing if required.

Digital photography equipment

The following is a inventory of my current equipment:

Camera

• Nikon D200 – the Nikon 200 is good enough - (see Canon 5D Mark II discussion in blog below)
• AF Nikkor 50mm 1.8 Lens – Nikon's cheapest and sharpest lens
• MC-30 Cable Release – essential for taking sharp photos and bracketing for HDR.
• 1x 8GB Card – no need to switch cards in the middle of a shoot (cheap as chips too)

Tripod

• Manfrotto Tripod 190 ProB - an entry level tripod that could be upgraded for added stability (but will do for now)
• Manfrotto 141RC Head - basic pan and tilt head that seems to work (could upgrade to a pano head but hard to justify the cost at this time)

Accessories

• Head torch – essential for early morning and twilight work
• Level – I use a small carpenters level from the hardware store (a good habit picked up from levelling the Sinar)
• Inverter – plugs into the car lighter socket and will charge camera batteries and laptop while on the road.
• Spyder2 Express – simple and foolproof color management for the screen
• 300w Inverter - Plugs into the cigarette lighter socket for charging laptop and camera batteries on the road
  

Computer & Software

• Laptop – I’d love to be a Mac user, but my work supply’s me with a Dual Core, 4GB Ram laptop running Windows XP.
• Maxtor 1TB HDD – 1TB used to seem a lot, but by the time six images are processed and the working files including NEF’s are backed-up you’re looking at 25GB for a single shoot.
• Printer – I still use my old trusty Canon SJ800 for proofing and send out to the labs for final prints.
• Adobe Camera Raw 4.0 – Have considered upgrading to Nikon Capture but have yet to justify the expense.
• PhotoShop CS3 – Still the default post-processing software.
• Autopano Pro – Best of breed open source photo-stitcher.
• FDR Tools - Easy to use photo realistic HDR image processor.

The beauty of this kit is that it all fits in one small backpack – including the laptop - no need for any specialist back pack, as I have a LowePro 170 AW shoulder bag that neatly fits into the backpack and protects the camera.

Large scale digital photography

My objective has always been to produce quality large scale prints full of detail and high resolution – the question has always been which scale and by using what process? With conventional DSLR images, I was always frustrated by the lack of detail and resolution of the image when compared with large format film images. Over the winter I’ve reconsidered my whole approach to large scale photography.

I found with the Sinar that I was spending so much time setting up shots, that I was missing the light. Further, the logistics and cost of large format film photography made me reluctant to go out and experiment as freely as I do with digital. While some of this frustration may be put down to my inexperience with the medium, I believe the images from Lake Mokoan using stitching to produce large scale images is a compelling argument to switch to digital.

the belt of venus, lake mokoan, originally uploaded by Andrew McDougall.

Lake Mokoan is three hours north of Melbourne just out of Benalla on the Hume Highway. The man-made lake is a reservoir for the surrounding farming district developed in the 1970’s by merging two existing swamps. As a result of the rising lake levels, a forest of red gum trees was drowned. Today the drowned red gum forests stand like sentinels within and on the edge of the lake. In November 2008 I took photos at both sunset and sunrise at the same location. I used a large scale digital image stitching technique that I’d developed back in Melbourne using the West Gate Bridge as a subject.

saltwater lake, westgate park, originally uploaded by Andrew McDougall.

Without wanting to enter into the film vs digital debate, here is what makes sense for me when considering the two formats:

Cost – My initial outlay for the Nikon D200 was more than the cost of purchasing a second-hand Sinar. However, even a second hand set of LF lenses (90mm, 150mm, 210mm) is considerably more than the equivalent my set of digital lenses (18-200mm zoom and 50mm prime). With digital there is also an initial cost in purchasing a computer and software. Unless one is going to stick to the darkroom (and I never was), this cost is also shared with the large format film scanning workflow.

So although the outlay for the equipment is roughly the same, it is in the ongoing costs that digital is so efficient. With the cost of storage coming down all the time it is possible to purchase an 8GB compact card for less than $50 and 1TB HDD for less than $200. In contrast the price of large format film consumables is expensive:

• Film: $75.20 for 10 sheets of Fuji Velvia 100 @$7.50 per sheet
• Development - @6.60 per transparency
• Scanning - 2040 PPI 16Bit 500MB @$25 per file
  

If every shot is bracketed (even the pros do this) at 1/3 stop then a single image costs the following:

3 x sheets @$7.50 = $23.50
2 x development @$6.60 = $13.20 (this is assuming that the 0EV shot is developed first and a bracketed exposure is developed second only if required)
1 x scan @$25 = $25
Total = $61.70

Considering that there are also shots that do not make it to the scanning stage, for each failed image there is the following cost:

3 x sheets @$7.50 = $23.50
2 x development @$6.60 = $6.60

Total = $30.10

At Mokoan I processed six images, assuming that I could have set-up the Sinar fast enough to capture the rapidly changing light, this would have cost me $370.20 for the developed images and at least another $180.60 for six rejected shots, assuming for every processed image there is a rejected image (in reality the ratio is likely to be higher – 3:20 images at sunset) - for a total of $550.80.

For a commercial photographer working on behalf of a client this may cost be justified. For a hobbyist, or semi professional these costs border on delusional to justify. This also assumes that the client is willing to wear the additional costs for the perceived increase in quality of the large format film image.

Time – On top of the costs above there is also the time to process the large format film image and the digital stitched image. The elapsed time of capturing the image, to develop, to scan and post process can run into weeks. While the capture of multiple digital images and the stitching time consuming, the lapsed time is only a fraction of the film camera. In addition there is also the logistics of buying the film, taking the film to the lab for development and scanning (may require repeat visits if bracketed images are required as well)

Flexibility – In landscape photography at twilight the light can change in an instant, the view camera is not flexible enough to be relocated – these two photos were taken within minutes of each other – there’s no way I could have setup the View Camera in such short time, focussed, loaded the film and bracket shots.

These two images were taken within minutes of each other:

sunrise and trees, lake mokoan, originally uploaded by Andrew McDougall.

first light, lake mokoan, originally uploaded by Andrew McDougall.

Quality – There’s no double that a well exposed and scanned 4x5 is a quality image, the degree of detail in the film image is superior. However, I believe the combination of a fixed focal length lens (Nikkor AF 50mm f1.8), mirror up, tripod, RAW and cable release can come close to matching the output of a scanned 4x5. A double layer stitched image of 12-22 exposures equates to a 500MB file which is roughly equivalent to a 4x5 scan. Both formats have limitations and compromises have to be made. I’ve made the decision to sacrifice some detail of the large format film image for the flexibility and cost savings of the large format digital image. If I had unlimited time and money I might consider the large format film process over the digital process - but I’d still mourn the loss of images due to the setup time required for the View Camera. Others may choose differently.

Exposure – This is where digital comes in to its own. I can check the exposure on location using the RGB histograms and adjust the shutter speed as required. With the Sinar I had to bracket at least -1/3 and +1/3 stop as a minimum, that’s an extra cost for film and processing and still not have any guarantee that I nailed the exposure. With digital I point the camera at the brightest part of the image, usually the sky, and adjust the exposure until the highlights to the right in all three RGB histograms are just short of being clipped. If the sun is within frame I allow this part of the image to be clipped so as not to clog the shadows to the left.

By using the 50mm lens I ensure the sharpest image with no barrel distortion or wide angle diffraction. Up until a few months ago, frustrated by the Sinar’s inflexibility, I was itching to upgrade to a Canon 5D Mark II in the pursuit of sharpness and resolution. Now I’ve discovered digital stitching I’ve realised that my Nikon D200 is “good enough.” At 15MP per raw file my Dual Core PC is already struggling with the 1/2GB PSD files. The Nikon D200 is a 10MP camera as opposed to the 22MP of the Canon 5D Mark II; I’m afraid to think how large the resultant PSD file would be. Besides, I love the metering of the Nikon's and the usability.

There is no doubt the process of using the View Camera has taught me valuable lessons that I would never learned using digital. The manual process made me think in new ways about the relationship of aperture, ISO and shutter speed as well as focus. In many ways it has simplified my digital photography technique, I now use manual process for both exposure and focus control. As I’m shooting RAW, I can forget about filters, white balance, colour and saturation at the time of capture. With digital I also have the option of capturing HDR images that would never be practical with film.