Here's my workflow. I'm not presumptive enought to say that this is "the" way to do it, but it's what I've become comfortable with. Not trying to be patronising either, but I'll add a step by step walk through.
I'm not quite happy with either finished article, and I think I'm going to have to repeat this process twice, creating two images, 1 for the sky/reflections (bringing back a little blueness) and 1 for the buildings, before blending together in PS at the end.
Part 1 - the shooting:
After setting up the tripod with the camera in portrait orientation, I focussed on Infinity Tower (the twisty one) and then locked the focus using the switch on the lens.
In camera, I switched White Balance to Manual, and then dialled in 2860K (I think). I find that 2800-3000K generally hits the spot for nightscapes in Dubai, but it will depend on the type of lighting in the shot. Obviously you can just shoot in RAW and then adjust the WB afterwards, but I wanted to avoid this batch process at source.
I was set up at ISO200 and dialled in f16 at 70mm. I switched the camera to Manual exposure mode and adjusted the shutter until the metering indicated -1EV on the LCD. Generally I find that at night in Dubai this gets the right exposure for me, but obviously it depends on what is in the frame. The was resulting in around an 8 second exposure.
Using the remote shutter release, I shot a frame to view how things were looking (to see if I needed to get the shot steadier by using mirror lock up or shutter delay)
Once happy with all of that, I set up bracketing in camera (-2, -1, 0, +1, +2EV). This would give me exposures of 2", 4", 8", 15", 30". Give it a try at night, the longest shutter speed will expose your scene better (in this case the buildings) but blow the highlights (in this case all of the lights and the areas in the vicinity) out of the water. The shortest shutter speed will still have blown highlights with point light sources, but the vicinity will be better exposed. As mentioned by lagnificent, using the -1 and +1 exposures is debatable, but I find that is generates extra microcontrast and detail when blending - the trade off is of course extra time processing.
Then it was a case of starting on the left of the image, shooting the bracketed shots and moving to the right, each time insuring a 20-30% overlap to help with stitching. Once in the can, it's onto part 2.
Part 2 - the processing
I personally used a couple of shortcuts at this stage, mainly to try to reduce file size as I don't have a lot of processing power under the hood. For the best detail possible, I could have opened every RAW file, processed to 16-bit .tif files, and used that as a starting point for the stitching. As each file would then be around 70Mb, that was a no-no. Therefore I jumped straight to the stitching software and used the JPG files straight from camera (remember I used manual WB in camera, so all my .jpg images will be exposed correctly).
After opening PTGui, I imported the 45 .jpg files. PTGui automatically recognises that there are multiple exposures of each image and allows for this. During the prompt process when aligning the panorama I requested processing as Exposure Fusion rather than HDR. Exposure Fusion is the lightest form of "HDR" and creates more realistic results.
After alignment, I then had to go into each image to create my vertical markers to make all buildings vertical. It's easiest to do this with the brightest exposure of each set of 5, and you don't have to do it to each image in the brackets, just the one. Seeing as there are lots of buildings with vertical lines, it's easy to add vertical references by plotting down the sides of buildings, windowlines etc.
Once this process was done with all pictures in the stitch, I then optimised. All the verticals were now pretty much vertical, but as the panorama was such a wide angle (around 110 degree field of view) there was still a slight fish eye effect. Choosing Cylindrical Panorama solved this instantly.
Then in PTGui you can optimise your exposure and blend. Finally you can hit the button to save your image as a 16 bit .tif file.
That might all sound alien, but there's a free trial of PTGui with handy FAQ and tutorials on their website - it takes a bit of getting used to but creates good results.
Finally it was a case of opening in Photoshop, and tweaking a couple of curves and I was done.
As I said at the beginning, I'm not entirely happy with the results. Other options which will give me a better finished article are:
Create 5 panoramas (each using 9 pictures at differing exposures). e.g. one panorama at -2EV, one at -1EV etc. After creating all 5 panoramas in PTGui I could then use Photomatix Pro to create a better Exposure Fusion blend.
Create 1 panorama, tweaking the sky and reflections in one copy and the buildings in another, before blending in Photoshop.
So there you go - I'd recommend trying out the trial software and see how you get on.
Last edited by dubaiphil on Sun Jun 12, 2011 6:43 am, edited 2 times in total.