gfish: (Default)
gfish ([personal profile] gfish) wrote2008-02-12 03:01 pm

Camsell Mountains

I finally got around to playing with the autostitch tool. (Developed here at UBC, using genuine homegrown SIFT features!) I've had some panorama shots sitting around in my image database for years, so it has been fun finally putting them together.





(Linked to the full image, which is hyuj. But worth it.)

These are the Camsell Mountains, as seen from our day 9 campsite on the Mackenzie. This was shortly after the Camsell Bend, where we finally started to head north. Despite how nice it looks, we had spent the last day and a half fighting absolutely horrible headwinds . We had triumphed over the elements, made our target distance, were finally racking up some real latitude, and even found a fairly nice campsite. Morale was high.

I'm really glad I bothered to get these pics. That was a good night. That was a satphone night, and talking to loved ones while taking in this vista is a powerful memory.

[identity profile] adularia.livejournal.com 2008-02-12 11:28 pm (UTC)(link)
Hey, I remember that campsite! After Axe Point, it was one of my favorites.

Remember the the 270-degree turn against the current, on our first mooring attempt?

[identity profile] airlinepilot.livejournal.com 2008-02-12 11:30 pm (UTC)(link)
Nice pic. No Mac support yet. :(

[identity profile] koruja.livejournal.com 2008-02-12 11:31 pm (UTC)(link)
I find it hilarious that the sample image on the website shows a guy in a red jacket among the original images who is then melded into a red splotch in the final stitched imaged.

[identity profile] gfish.livejournal.com 2008-02-12 11:36 pm (UTC)(link)
The trick of splitting high and low frequency components works pretty well, but it isn't perfect. And I suspect the samples on the page are some of the earlier output -- who is going to bother updating the webpage? Ah, academia. :)

[identity profile] gfish.livejournal.com 2008-02-12 11:39 pm (UTC)(link)
Looking at it closer, the red blob is actually a backpack that is in some of the source images for that area, but not all. That guy must have been walking around, rearranging objects in between the pictures being taken. Not much an algorithm can do about that.

[identity profile] gfish.livejournal.com 2008-02-12 11:41 pm (UTC)(link)
I think that was the day 5 campsite, right after the river narrowed.

Also good times: those currents around the rock face after Wrigley. Whee!

[identity profile] koruja.livejournal.com 2008-02-12 11:43 pm (UTC)(link)
I kind of read it visually as a campsite with two guys, one standing up in the red jacket and the other kind of leaning up against some rocks (yellowish/orange and purple jacket). Stupid people not knowing how told hold still for a few dozen photos.

[identity profile] gfish.livejournal.com 2008-02-12 11:45 pm (UTC)(link)
I could probably find the fullsize original images on a file share somewhere, if you really care. ;)

[identity profile] neuro42.livejournal.com 2008-02-13 12:08 am (UTC)(link)
http://hugin.sourceforge.net/

[identity profile] adularia.livejournal.com 2008-02-13 12:16 am (UTC)(link)
Hm, you're right, the mountains weren't in sight that night. Just a lot of scraggly trees.

[identity profile] gfish.livejournal.com 2008-02-13 08:11 am (UTC)(link)
Ew, manual point labeling? We have the technology to automate the entire process! It's almost like random sourceforge people don't read computer vision journals or something.

[identity profile] neuro42.livejournal.com 2008-02-13 08:28 am (UTC)(link)
No, automatic. Manually adjustable if it gets it wrong.