On Thu, Apr 7, 2011 at 8:13 PM, Christopher Allen <cpcal...@gmail.com> wrote: > On Apr 7, 2011 6:15 PM, "Yclept Nemo" <orbisvi...@gmail.com> wrote: >> >> RAW images are in linear color space so hugin would >> not have to reverse-calculate the response curve applied by >> ufraw/dcraw. > > Is that generally true? I don't know much about raw processing (or indeed > about image sensor electrical behaviour) , but I wouldn't have guessed that > light input -> raw pixel values would be necessarily or even typically > linear.
I'm pretty sure it's accurate: http://www.luminous-landscape.com/tutorials/expose-right.shtml The gist: 2x light = 2x voltage = 2x pixel intensity. Since human vision is logarithmic-ish, perceiving a greater range of luminance across darker values, a centred exposure fails to utilize the upper-half of the camera's sensor. Anyway my point is that since RAW images represent actually the image recorded by the sensor, reading directly from these files makes moot the need to apply a camera response curve or optimize for white-balance. >> And I have an additional question: >> >> I have a lightpost running vertically straight through the center >> section of an overlapping area between two images. There are 40 >> control points scattered across these two images, and I am sure all >> control points are accurate. Furthermore each control point can be >> optimized to an error of less than 1px. So I am stumped as to why, in >> the stitched output, the lightpost is misaligned by at least 20px. >> Does anyone have any suggestions? > > Well, if it were my panorama it would be because of parallax from > hand-holding the camera, but I'm assuming that you're using a calibrated > panorama head. > > (And a much better lens: my 18-200mm Nikkor has enough variation in FoV due > to aperture and focus changes there I can't get a good fit when processing > image-stacked panoramas without optimizing (unlinked) v too.) Yes I did manage to borrow a panoramic head; even without out, since the lightpost is about 50-75 feet distance I doubt parallax would cause any problems. Interestingly enough I found that since the panoramic head is so large vertically and lacking a clamp, at certain angles it must have acted as sail and subtly rotated - a few of my brackets stacks are misaligned. I use a cheap Canon EFS 18-55mm lens which I calibrated specifically for 28mm Anyway, is it possible to produce a perfectly aligned 360 degree panorama? I've been trying really hard - many different strategies - and am unable to get rid of artifacts. I find it interesting how radical and under-realized a job the blender does - I wish hugin would allow a visualization of the seams (graph paths) used by enblend: Different strategies: 1] Let hugin/CPFind pick points. Many many points. Many don't actually correspond to identical features despite high correlation. Images are drastically unaligned, yet enblend does a very good job. Very few artifacts (2-3) with high error (10-50px perceived) 2] Let hugin/CPFind pick points, then manually subtract. Never finished. Too many points to filter, plus optimization strategy wants to get rid of my points (all my points -> 100+px error distance) 2] Pick points myself. 20-40 per stack. Precise + high correlations. Images are very well aligned, yet many minor artifacts produced. 10-20 artifacts with error of (2-10px) 3] Pick Points + Use straight lines. Perhaps I don't know how to use straight lines. In any case, whereas normal points average error of <1px, line CPs average error of 10-20px. 20-30 artifacts with error of 2-10px, visually errors are less noticeable than in previous strategy. 4] Pick Points + Use vertical lines. Work in Progress. Anyone have tips? This is really frustrating. -- You received this message because you are subscribed to the Google Groups "Hugin and other free panoramic software" group. A list of frequently asked questions is available at: http://wiki.panotools.org/Hugin_FAQ To post to this group, send email to hugin-ptx@googlegroups.com To unsubscribe from this group, send email to hugin-ptx+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/hugin-ptx
