I've been provided with two images of 16-bit unsigned pixel type.
When i'm using the clip raster tool in ArcMap the output raster pixel type changes to 32-bit. Neither in the dialogue box, nor in the environments have I found the option to set it manually. I set format to TIFF since it's the most widely supported one. I know that this format also supports 16-bit unsigned pixel type.
Any ideas why such thing happens and how can I change it?
Found this on an ESRI forum:
simo3919 says:
Clipping with geometry will assume that you will add a NoData value. If you specify a NoData Value, then you will not have to promote the pixel depth. Be aware though, if that value already exists in the data, all those pixels will now become NoData. The safer route is to promote your pixel depth.
Why is it safer? If you specify a NoData value that is already a valid pixel value in your raster, then all those pixels are now NoData values as well. However, promoting the pixel depth will create a new value to act as the NoData value.
Why do we need to promote pixel depth? An 8-bit raster can only hold 256 values. When no NoData value is specified, a new pixel value is added to ensure that no valid values are overwritten. Therefore, this means you will have 257 values now – hence the need to promote the data to the next bit depth level (16-bit).
Not my answer, found it here http://blogs.esri.com/esri/arcgis/2009/05/06/better-raster-clipping-options-in-arcgis/
Hi Damian,
I'm not sure exactly, but I suspect Arc is set up to output the lowest possible bit depth among 8, 16, 32 and 64, etc., that still covers the range of values in your data, in order to conserve disk space. Or perhaps some tools are just set by default to use the large 32-bit format.
I know of two open source tools that will let you specify a bit depth in raster creation: 1) GDAL, specifically the GDAL Translate command-line tool (http://www.gdal.org/gdal_translate.html) will let you copy the raster, and if you use the -ot flag, you can set the output pixel depth to whatever you need. I attach a screenshot where I copy a raster into an Int32 version. You could use Int16 or UInt16, depending on whether you need signed or unsigned values.
Also, SAGA GIS has a tool called Change Data Storage (see http://www.saga-gis.org/saga_module_doc/2.1.3/grid_tools_11.html) that will copy a raster into whichever format you choose. I used this tool just yesterday :) You can access it through QGIS, if you've installed SAGA as well as QGIS (automatically done in the standard Windows installer). It's also accessible via command line, like GDAL.
Cheers :)
Found this on an ESRI forum:
simo3919 says:
Clipping with geometry will assume that you will add a NoData value. If you specify a NoData Value, then you will not have to promote the pixel depth. Be aware though, if that value already exists in the data, all those pixels will now become NoData. The safer route is to promote your pixel depth.
Why is it safer? If you specify a NoData value that is already a valid pixel value in your raster, then all those pixels are now NoData values as well. However, promoting the pixel depth will create a new value to act as the NoData value.
Why do we need to promote pixel depth? An 8-bit raster can only hold 256 values. When no NoData value is specified, a new pixel value is added to ensure that no valid values are overwritten. Therefore, this means you will have 257 values now – hence the need to promote the data to the next bit depth level (16-bit).
Not my answer, found it here http://blogs.esri.com/esri/arcgis/2009/05/06/better-raster-clipping-options-in-arcgis/
- Badges
- Science topic
- Similar topics
- Computer Science
- Images