[hugin-ptx] Re: Control point generation in difficult conditions - some success
I also tried a similar method with GIMP on your photos using other filters such as unsharp mask with the same substitution idea in mind. Match points were found, but they tended to be on the upper elements that would introduce parallax errors when used as match points. It seems to me that for these photos you need to match to elements occurring only in the narrow band in the middle of each image that is directly below the camera while also cropping each image in Hugin so that only that middle portion is used for the final image. It looks like reliable matching would be possible in these images if the auto process were to compare large areas for matching instead of small areas. Perhaps the processes have a variable to control this. Allan On Fri, Sep 25, 2009 at 3:56 AM, Oskar Sander wrote: > I made a workaround with some sucess that I have yet to try on a bigger > scale than on two pictures. Have a look at this project [1]. > > I pre processed the photos using Adbe lightroom. I made quite radical > adjustments: >Sharpening 91 at radius 1.0 >Exposure +0.4 EV >Contrast +100 > > Doing this, APC finds 11 CP instead of 0. Of which 2 were completely off, > but they are excluded automatically if ransac is used, see [1]. > > To the human eye (mine:-) the selected CP doesn't use the obvious and most > prominent features, but this is still good progress! > > My idea is to run CP finding on preprocessed images, and then substitute > these for the originals before the stich step. > > I'd like to understand the APC inner working better though, maybe there are > potential improvements. I've been peeking a bit in the code of APC and the > LoweFeature detector. It seems to look for contrast rich areas, is it using > the colour channels separately or luminance? Any suggested reading? > > Dale, there is a reference to layout mode in the parameter list, but that > may be one not-yet-working feature of APC (right?) > > > > [1] > http://sites.google.com/site/thebigo1984/mosaiks/UW_low_contrast_workaround.zip?attredirects=0 > > > > --~--~-~--~~~---~--~~ 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 -~--~~~~--~~--~--~---
[hugin-ptx] Re: Control point generation in difficult conditions - some success
Hi Oskar, did you also try panomatic? It might work better than autopano-sift-c. ciao Pablo Oskar Sander schrieb: > > > My idea is to run CP finding on preprocessed images, and then > substitute these for the originals before the stich step. > > I'd like to understand the APC inner working better though, maybe > there are potential improvements. I've been peeking a bit in the > code of APC and the LoweFeature detector. It seems to look for > contrast rich areas, is it using the colour channels separately or > luminance? Any suggested reading? > > > > Hmm, no it did not improve much when tried en-masse unfortunately. > > The places where controllpoints are found ar typical protruding features > that are better lit man more contrasty than the deck, but these does > not work for optimization as they are not in the deck-plane. > > Is there an APC constant or parameter to lower the threshold of APC > critera, that could be tweaked in order to experiment with the selection > criteras if the images are too soft in general? > > > -- > /O > > > --~--~-~--~~~---~--~~ 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 -~--~~~~--~~--~--~---
[hugin-ptx] Re: Control point generation in difficult conditions - some success
> > My idea is to run CP finding on preprocessed images, and then substitute > these for the originals before the stich step. > > I'd like to understand the APC inner working better though, maybe there are > potential improvements. I've been peeking a bit in the code of APC and the > LoweFeature detector. It seems to look for contrast rich areas, is it using > the colour channels separately or luminance? Any suggested reading? > > > Hmm, no it did not improve much when tried en-masse unfortunately. The places where controllpoints are found ar typical protruding features that are better lit man more contrasty than the deck, but these does not work for optimization as they are not in the deck-plane. Is there an APC constant or parameter to lower the threshold of APC critera, that could be tweaked in order to experiment with the selection criteras if the images are too soft in general? -- /O --~--~-~--~~~---~--~~ 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 -~--~~~~--~~--~--~---
[hugin-ptx] Re: Control point generation in difficult conditions - some success
I made a workaround with some sucess that I have yet to try on a bigger scale than on two pictures. Have a look at this project [1]. I pre processed the photos using Adbe lightroom. I made quite radical adjustments: Sharpening 91 at radius 1.0 Exposure +0.4 EV Contrast +100 Doing this, APC finds 11 CP instead of 0. Of which 2 were completely off, but they are excluded automatically if ransac is used, see [1]. To the human eye (mine:-) the selected CP doesn't use the obvious and most prominent features, but this is still good progress! My idea is to run CP finding on preprocessed images, and then substitute these for the originals before the stich step. I'd like to understand the APC inner working better though, maybe there are potential improvements. I've been peeking a bit in the code of APC and the LoweFeature detector. It seems to look for contrast rich areas, is it using the colour channels separately or luminance? Any suggested reading? Dale, there is a reference to layout mode in the parameter list, but that may be one not-yet-working feature of APC (right?) [1] http://sites.google.com/site/thebigo1984/mosaiks/UW_low_contrast_workaround.zip?attredirects=0 --~--~-~--~~~---~--~~ 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 -~--~~~~--~~--~--~---