Tuesday, February 8, 2011
Supersized Panorama Stitching
Brandon Riza’s panoramas are crafted in the spirit of an Ansel Adams landscape. In each one, the mountains are larger than life, and the overall image is richly detailed and built from a series of high-resolution frames.
After importing images into Lightroom and running lens-vignetting and chromatic-aberration corrections, I pull up dark areas as needed with the tone curve or histogram tools. I then apply these corrections to all the images in the sequence and export them all as 16-bit TIFF files. Sharpness adjustments aren’t made until after the final pano is created and then only on individually downsized files.
The frames then go into Autopano Giga, and I launch the initial detection. I’ve found this software to be faster and more accurate with more options than other stitching applications. It allows for multiple tweaks for fine-tuning the end result. Once the initial stitch is calculated in Autopano, I’ll enter the Editor and make any necessary adjustments. It displays the field of view of the pano calculation and gives you an idea of default stitch accuracy. A high RMS means a bad stitch (more common when the frames are shot with a wider lens) while a low RMS means a stable stitch. These RMS values are the sum total of the accuracy of the control points the software created automatically when the initial detection was launched.
It can take anywhere from a few minutes to an hour to build the pano solution. This is the initial number-crunching stage. I usually run about five to 15 panos through the software at a time. I process hundreds of images in Lightroom, export them into their organized and named folders, and feed those folders full of frames into Autopano and run the initial detection on all the panos simultaneously while I’m away. That part of the process can take hours, days, sometimes even weeks. And that’s not counting the renders—which can take minutes to weeks of straight number-crunching. I’m typically dealing with enormous output files, in the 20,000- to 30,000-pixel-wide range, at 16 bits per channel, and the file sizes easily can exceed a gigabyte for a single render.
Once the renders are complete, I open them in Photoshop, crop as needed (though I normally crop the pano data), add my position and description text data as layers, add my saved metadata information and save the final as a 16-bit-per-channel LZW-compressed TIFF file.
Minimize Finishing Touches
I don’t like to do much more to my photographs. I think overprocessed HDR images look more like an artistic interpretation than an accurate reproduction of what the eye was seeing at the moment the frames were captured. I’m not saying this is bad; it’s just not my particular style. I prefer to represent reality, not artificially enhance it based on my own artistic interpretation. I strive to get as close to reality as possible with the tools that I’m using, then correct any aberrations those tools introduce.
To see more of Brandon Riza’s photography, visit brandonriza.com.
Page 3 of 3
Get 11 Issues of Outdoor Photographer for only $14.97!
That's 77% off the cover price!