Project Journal: Dali Museum Stitch (Surrender)

After a full day of working on this shot, five more stitches and at least 8 hours of control point and ghost removal I have succumbed to the fact that I made too many mistakes when shooting this image. I will publish a final crop of what made it, but this will likely never make it to print and I doubt I upload a full size version. Chalk this one up to a learning experience.

Project Journal: Dali Museum Stitch (cont.)

After initial reviews I have discarded all but two brackets of this piece, so I'm now stitching only 126 images. Removing those layers did mean I had to re-work a lot of control point links. I've also had to spend a lot of time trying to hide HDR ghosts. All in all about 4 hours of work, and just over 100 CPU hours of stitching (three more stitches). I've found several flaws that may not be fixable, and I'm still trying to hide ghosts. I've learned a good bit about trying to shoot in high traffic areas, and discovered the hard way that parallax is much more important indoors. While some crop's may be cool, and I'll likely still upload something, I'm not going to promote this one too much. I think I need some more practice on indoor work before I offer commercial real estate services, maybe the Basilica ?  

Project Journal: Dali Museum Stitch

Having narrowed the stitch down to 335 images, it took a bit to find the extra four frames (20 images) that were duplicates. The initial control point recognition showed me just how much work this would take, and after about an hour of control point tuning I ran the first stitch. I knew this wasn't going to yield a good product, but wanted to see where the focus needed to be. The new box made short work of this job, rendering in less than an hour. Over the course of the evening I spent about 3 more hours in control point tuning and ghost removal and ran three more stitches. Late in the evening I flattened a rendered image and loaded into a pano tour to explore. This is going to be a cool picture, here's a teaser: