On 14 Dez., 21:41, Matthew Gates <matthew...@gmail.com> wrote: > The DSS tiles are split from large, high resolution photographic > plates which are themselves subject to distortions due to the optical > properties of the telescope used. I can find a detailed description > of the distortion if necessary. It's not a standard barrel distortion > though.
It looks like that isn't the issue anyway and you have control of that bit of the distortion. When I, sloppily, said distortion, I meant the variations in brightness and colour. > We chop these plate images into slightly overlapping tiles, run them > though some programs which remove the plate specific distortion > preparing them for the toast projection method: > > http://www.worldwidetelescope.org/docs/worldwidetelescopeprojectionre... Excellent projection. I'll post a link to it to this group. maybe it'll inspire someone to utilize it for panorama output - if MS hasn't patented it? > Stellarium reads these tiles, knows about the toast projection and > translates this into sky coordinates "unwrapping" the tiles onto the > sky. Stellarium itself can project the curved sky onto the screen > with multiple methods, but this is not important for processing the > images. Now I'd like to know if the 256*256 tiles you have are already in TOAST projection? This would have warped the initial, probably concentric, vignetting into a more complex shape. Wouldn't it be a better idea to do the processing on the source images, before chopping them up? Do you have access to them at that stage? I am only guessing, but my feeling is that the image processing needed to make them blendable nicely could be done without having to take into account every small detail - it might suffice to define the procedure needed at a low resolution and then just let it trickle down to higher resolutions. If you were to try to model the vignetting on lots of small tiles, trying to undo an effect at the source image level, you introduce much more potential error than working on the source images straight away. But you may get away with figuring out the vignetting and mabe other problems in the source images by looking at lo-res versions of them and still come up with parameters suitable to treat them down to the highest resolution. > There are many factors here. Vignetting is probably a large part of > it. Individual tiles may come from close to or far from the optical > axis of the telescope which was used to create them, so if it's > possible to have the center of the vignetting compensation be far > outside of an individual image, that might work well. The exact way > the vignetting affects the image is probably not quite the same as > with normal camera lenses, but I dare say it is well documented and > details can be found if necessary. If it goes beyond vignetting, it is probably beyond hugin's scope. Hugin's workflow is - identify overlaps - make the images compatible geometrically and photometrically - blend them nicely The idea is to transform them so that they become as compatible as possible and then let the blender decide where precisely the seam should be put - but ideally, if they overlap, sections that are covered by two images should look identical on the warped versions of the two source images. So hugin could help you with making the images compatible, but it is really quite specialized insofar as it is geared towards photographs with the usual flaws that photography produces, which is reflected in the transformations it can perform. > The original plates were taken over a long period of time, so there > will be plate-to-plate variations which are much harder to compensate > for - atmospheric conditions when an individual plate was exposed such > as dust and water vapor in the atmosphere, altitude angle and so on, > scattering of light from bright objects such as the moon and planets > and so on. This is of course a fine line. How much emphasis do you put on the authenticity of the individual source images and how much on the experience of a continuous overall display? If you think, for example, of google earth, they seem to have chosen to take the satellite images pretty much as they are and accept that there are discontinuities at the borders. Maybe it would be acceptable to do that from some level of resolution - like you could have the lower-resolution tiles more modified towards a smooth overall display, changing emphasis to fidelity to the original images as you step up the resolution. The pyramid approach should give you that option. Also, you have stellarium as a reference. You could simulate every tile with stellarium, which should give you an idea about at least some of the characteristics the image should have at in that area, like average brightness (just a quick shot, the statistics could be much more involved). Maybe you could use these data to brighten/darken the corresponding real tiles. > The images can be downloaded, but it's hard to know which ones are > adjacent at a given resolution. I will try to research this. Great. Looking forward to an interesting investigation. with regards Kay -- 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
