I have a simple model that calculates the NDVI from bands 3 and 4 of a raster dataset. The input bands are 8-bit unsigned, whereas the default output NDVI is signed 32 bit.
How can I specify the pixel depth for the output NDVI as 8 bit unsigned?
-
Can this be controlled with an environment setting? – L_Holcombe Feb 12 '13 at 15:43
-
I haven't found a way to specify the pixel depth in the environment settings. – Aaron♦ Feb 12 '13 at 16:03
-
Does the calculator exhibit behavior that changes the pixel depth? – gm70560 Apr 30 '13 at 21:38
-
No as this is part of a model so I need to keep the image with the original 8 bit, do the calculations, convert it into 32bit and do some other calculations and then output it. The problem is to keep it at 8 bit for the first set of calcs. – John Apr 30 '13 at 22:03
-
Would you be able to edit your question to include this clarification to it, please? – PolyGeo♦ Jul 23 '13 at 23:12
The result from NDVI will be continuous (i.e. decimal) values between -1 to +1, therefore the raster must be able to store these values, and will use signed pixel depth. If you truly want 8-bit unsigned, you will need to adjust the expression in the raster calculator by linearly scaling to values between 0-255 and then applying the int() function on the entire calculation. However, this still does not guarantee that the output will be 8-bit unsigned. Perhaps try Copy Raster on the scaled raster with the optional pixel_type parameter in your model to help with this.
After running Raster Calculator, use the Copy Raster tool with the pixel_type parameter set to 8_BIT_UNSIGNED, as shown below.
