I have some issues with orthoresampling images in ERDAS Photogrammetry (formerly LPS). I've set up two different image blocks (including orientation and camera paramters) and created multiple digital surface models by semi global matching to use for orthorectification. These DSM have been converted to raster data (.img) and everything seems fine so far.
One Block is 4-channel imagery (RGBI), one is 3-channel (RGB). Both have a 8-Bit depth.
Now i've tried the Ortho Resample Tool to create Orthophotos from my arial imagery. Somehow the radiometric resolution seems to shrink with the process. I'm aware of changing grey values due to resampling, but how can the aviable value range shrink?
Example before orthorectification:
Example afterwards:
The geographic region shown in the output orthophoto is limited to a subset of the original image, so different greyvalue distribution is fine. But those masses of very low values and the limited range (0 - 206, different in every image, for example 14 - 222 occur as well) seems very wrong to me. How can this happen?
I've tried different dataypes (.tif and .img), different DSM and different resampling types (Nearest Neighbor and Bilinear). The outcome looks slighty different but the problem remains.
any ideas?
أكثر...
One Block is 4-channel imagery (RGBI), one is 3-channel (RGB). Both have a 8-Bit depth.
Now i've tried the Ortho Resample Tool to create Orthophotos from my arial imagery. Somehow the radiometric resolution seems to shrink with the process. I'm aware of changing grey values due to resampling, but how can the aviable value range shrink?
Example before orthorectification:
Example afterwards:
The geographic region shown in the output orthophoto is limited to a subset of the original image, so different greyvalue distribution is fine. But those masses of very low values and the limited range (0 - 206, different in every image, for example 14 - 222 occur as well) seems very wrong to me. How can this happen?
I've tried different dataypes (.tif and .img), different DSM and different resampling types (Nearest Neighbor and Bilinear). The outcome looks slighty different but the problem remains.
any ideas?
أكثر...