Interpolate Fill option in Rasterize
Interpolate Fill option in Rasterize
Am I missing something or shouldn't the Interpolate Fill Empty Cells option in Rasterize do exactly that? Instead it is filling the cells with one value.
Re: Interpolate Fill option in Rasterize
The interpolation can only be done inside the cloud "convex hull". Outside of the convex hull, it's not possible to interpolate values. Therefore we use the "Fill with" value specified below the option for these cells.
See http://www.cloudcompare.org/doc/wiki/in ... mpty_cells
See http://www.cloudcompare.org/doc/wiki/in ... mpty_cells
The biggest limitation is that it's limited to the convex hull of the non-empty cells. Therefore a custom value must still be specified (in the field below the 'Fill with' drop-down list) to fill the remaining empty areas (if any).
Daniel, CloudCompare admin
Re: Interpolate Fill option in Rasterize
Thanks Daniel! I was actually talking about the interpolation inside the convex hull not working properly, but I updated my Cloud Compare to the latest version and it is working now. Thanks for the quick response.
Re: Interpolate Fill option in Rasterize
Daniel,
This was super helpful, thanks! Knowing that the "fill value" is used for all empty pixels outside of the convex hull helped me understand why it looked like interpolation wasn't working, even though it was.
Any chance it'd be easy to include "keep these cells empty" as a fill option in addition to a numeric value? This would have been helpful for my particular use case. Instead, I made the fill value -999, then used R to set the values of all the cells equal to -999 to NA.
In case it's helpful for anyone else:
I used the CSF algorithm to classify ground/non-ground points, then wanted to rasterize the ground points to generate a digital terrain model (interpolating in between the ground points to get the terrain elevation underneath the non-ground points).
The reason it appeared that the interpolation method for filling empty pixels didn't work is because my study area is slightly rotated from due north. So there were empty cells within the study area (underneath the non-ground points) and these were the ones I wanted interpolated. But there were also empty cells in between the edge of my survey area and the rectangular bounding box that surrounded my study area. So when the interpolated raster was rendered, the empty cells between the edge of my survey area and the bounding box were depicted as all blue (because they were all filled as 0) and the empty cells that were interpolated appeared all red (because the range of real values for my terrain model is about 1100 to 1200).
This was super helpful, thanks! Knowing that the "fill value" is used for all empty pixels outside of the convex hull helped me understand why it looked like interpolation wasn't working, even though it was.
Any chance it'd be easy to include "keep these cells empty" as a fill option in addition to a numeric value? This would have been helpful for my particular use case. Instead, I made the fill value -999, then used R to set the values of all the cells equal to -999 to NA.
In case it's helpful for anyone else:
I used the CSF algorithm to classify ground/non-ground points, then wanted to rasterize the ground points to generate a digital terrain model (interpolating in between the ground points to get the terrain elevation underneath the non-ground points).
The reason it appeared that the interpolation method for filling empty pixels didn't work is because my study area is slightly rotated from due north. So there were empty cells within the study area (underneath the non-ground points) and these were the ones I wanted interpolated. But there were also empty cells in between the edge of my survey area and the rectangular bounding box that surrounded my study area. So when the interpolated raster was rendered, the empty cells between the edge of my survey area and the bounding box were depicted as all blue (because they were all filled as 0) and the empty cells that were interpolated appeared all red (because the range of real values for my terrain model is about 1100 to 1200).