Hi David, osgdem uses GDAL to do image reading and reprojection. Its the reprojection step that is the problematic one here, and stems from projection producing a black region around the image, osgdem tries to account for this by ignore all pure black pixels. If GDAL is returning some of images an not pure black due to sampling then these pixels won't be rejected.
In the rewrite of VPB I am considering doing reprojection using OpenGL and use an alpha channel to manage the boundary regions. Alternatively it might be possible to promote an RGB source image to RGBA and then get GDAL to reproject and then blend using the alpha channel. Robert. On 7/11/07, Riepl, David M <[EMAIL PROTECTED]> wrote:
Thom/All, I mentioned in my email, that I have verified in several viewers that the black stayed pure black after the reprojection step of GDAL. None of the pixels are even near to black as my sample is in a very desert-like area. Are you suggesting something later in the process that is caused by osgdem? I don't even know, is osgdem somehow running GDAL as a sub-process? David Riepl Visualization IG\DB Team >Training Systems & Services [EMAIL PROTECTED] 314.777.4741 -----Original Message----- From: Decarlo, Thomas R. (LARC-B702)[NCI INFORMATION SYSTEMS] [mailto:[EMAIL PROTECTED] Sent: Wednesday, July 11, 2007 1:54 PM To: osg users Subject: RE: [osg-users] Black seams visible in osgdem output David, Those black lines are probably the result of the resampling step. GDAL (and hence Osgdem) will, by default, consider the black borders around the images to be the background color and will ignore them when mosaicing the textures. However, when using the cubic spline the black background pixels immediately adjacent to the non-black image pixels can be blended with the image pixels and become a non-black color. The width of the non-black border pixels will depend on the size of the cubic kernel. To avoid the blending, try either - using nearest neighbor resampling. - use the cubic spline resample but then re-mask the output to trim off the edges of the output images that might be blended. HTH, Thom -- Thom DeCarlo mailto:[EMAIL PROTECTED] Sr. S/W Eng, VisSim -- NASA LaRC AvSP -- v:757-224-4089 -----Original Message----- From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of Riepl, David M Sent: Wednesday, July 11, 2007 2:41 PM To: osg-users@openscenegraph.net Subject: [osg-users] Black seams visible in osgdem output All, When I make a 2-D Map from satellite imagery and let it generate the maximum number of LODs required to reach the max resolution, I always see black seams/stripes/borders of my images used during the OSGDEM process. Here is my normal approach: I start with a folder of imagery...as an example let's say I have 12 geocells (1deg lat/1 deg long each) of imagery. Using gdalwarp (from FWTools), I will reproject all the satellite imagery from Geographic Lat/Long to my output projection of choice (usually TM, UTM or LCC). I use cubic spline resampling and output GeoTIFFs. I used to think this is where my problem lay, but I can open up the reprojected imagery in a viewer (like Imagine or GlobalMapper) and I can clearly see that the black edges around the terrain stayed pure black (0,0,0) after the reprojection. I can also verify that there is overlap between the images (3 pixels in my current case). So then I run osgdem on my folder of reprojected GeoTIFs with a command like this: Osgdem -t Y:\Test_10m_Reprojected -o test_sat_2d.ive -l 99 Most of the time the black seam does not become apparent/visible until I zoom into the highest LOD using the osgviewer. I am using osgdem from OSG1.2 on WinXP PC. HELP! David Riepl Visualization IG\DB Team > Training Systems & Services [EMAIL PROTECTED] 314.777.4741 _______________________________________________ osg-users mailing list osg-users@openscenegraph.net http://openscenegraph.net/mailman/listinfo/osg-users http://www.openscenegraph.org/ _______________________________________________ osg-users mailing list osg-users@openscenegraph.net http://openscenegraph.net/mailman/listinfo/osg-users http://www.openscenegraph.org/ _______________________________________________ osg-users mailing list osg-users@openscenegraph.net http://openscenegraph.net/mailman/listinfo/osg-users http://www.openscenegraph.org/
_______________________________________________ osg-users mailing list osg-users@openscenegraph.net http://openscenegraph.net/mailman/listinfo/osg-users http://www.openscenegraph.org/
