[HDRI] Photosphere 1.6 available

Mark Banas (lists) listmail at mab3d.com
Thu May 11 19:27:36 CEST 2006


On May 11, 2006, at 2:12 AM, Gregory J. Ward wrote:

> The panorama stitcher in Photosphere does a better job than most at  
> hiding gradient differences between images, so things like lens  
> flare are not the usual disaster.  Instead of seeing a seam or  
> strong gradient, you may notice an erratic line where the two  
> images are stitched together.  That's the way this algorithm looks,  
> I'm afraid.  Sometimes it looks better than a smooth blend, and  
> sometimes it's worse.  It really depends on the scene.

Hi Greg,

I tried out the latest function in Photosphere and was able to pretty  
quickly get good results from an LDR image set, but then I threw your  
new toolset a curve by extracting multiple rectilinear images (no  
lens distortion) from an otherwise happily stitched HDR original. (I  
did this because all my panos are shot with fisheye lenses, and I was  
just playing around...).

The results of this experiment are "mixed" to say the least. The HDR  
is retained (yay!) and out of the 4 image used there were minimal  
artifacts (this coming from a longtime Panotools/ Stitcher/ QTVRAS  
abuser), and only one seam was "erratic" and it seems to be in an  
area of high detail (buildings). You can see the results of this test  
(in JPEG-HDR) here:

http://www.mab3d.com/temp/ot3d/panotest1.jpg

I'll admit that this isn't a "fair" image, since only pinhole cameras  
(and in this case, Flexify 2) can create rectilinear images  
naturally, but overall it seemed to be a success. Yes, the interface  
for putting the pano together is nebulous and "well-hidden," but then  
its not supposed to be Hugin or AutopanoPro. When the rain stops, I  
can dig out the 18-55 "kit lens" and do some HDR tests with a "real"  
lens. (Maybe I'll go up to the rooftop putting green again...)

-Mark



More information about the HDRI mailing list