Followup to unanswered question - how to convert rasters to 1-band grayscale?

[StavrosC]

Hi everyone, I'm asking this again in hopes that someone can help.  Please don't hold the repost against me! 

I have hundreds of multi-band, georeferenced JPGs that I need to convert to single band geotiffs. All images are black and white, so I only want one band to minimize file size. 

Does anyone know tool that I can use to do this in a batch process? I've tried copy raster, mosaic to new raster, and composite bands. None of them will give me a single band raster, although copy raster produces a nice 3-band geotiff. Inserting a grayscale function through the image analysis window worked well but only for a single image at a time. 

Thanks so much for any thoughts! I know there is a way to do this....

[JeffreySwain]

Have you considered using the mosaic dataset and then applying the Gray Scale Function?  Once you have the grayscale, you can then export it.

[StavrosC]

Thanks for the response.  I've also looked into the grayscale function but have not been able to batch process the transformation of individual jpg grids to grayscale geotiffs. 

I created a sample mosaic of 5 of the 400 jpgs and inserted a grayscale function through the image analysis window.  I then used right click -> export data.  This successfully created a 1-band geotiff.  The problem is that this merges the images into a single raster, but I want the result to be a collection of 400 1-band geotiffs with the same extents as the initial 3-band jpgs, not one huge geotiff. Can you advise?  Maybe I need a Python script to average the three bands?

[StavrosC · Accepted solution]

I found a solution through modifying python script that was designed to calculate NDVI by performing calculations over raster bands.  My script is rudimentary, and it creates a temporary raster file for each of the three bands, then averages them and creates a new raster.  If you are trying to do the same thing, this script may help.  You will have to modify it to reflect your file names.


[INDENT]#Code is adapted from http://esriaustraliatechblog.wordpress.com/2012/03/14/using-python-to-calculate-ndvi-with-multiband-...
#Code is being modified to convert 3-band, georeferenced jpegs to single band geotiffs raster.

import arcpy, string

#import the environment settings and spatial analyst extension.  Import time so you can clock the geoprocessing speed.  Import glob to list file names. 

from arcpy import env
from arcpy.sa import *
import time             #Let's see how long this takes
import glob

# Overwrite pre-existing files so that temporary rasters can be copied over one another. 
arcpy.env.overwriteOutput = True

#Check out the Spatial Analyst extension.

arcpy.CheckOutExtension("spatial")

#Add your workspace here

env.workspace = r"I:\alb_aerials\1957"

#Create list of input and output file names
jpg_files=[]
for line in glob.glob("I:/alb_aerials/1957/*.jpg"):
  arcpy.env.overwriteOutput = True
  jpg_files.append(line)                                   #Add jpg file name to jpg file list
  input = line[20:]                                        #For my directory/file names, the file name starts after the 20th character.
  print input


#Create raster directory and specify inputs. It references the input you specified above
  result = input[:-4] + ".tif"
  B1 = input + "\Band_1"
  B2 = input + "\Band_2"
  B3 = input + "\Band_3"

#Define outputs ??? Note:  These will need to be deleted if you need to run this script again
  B1_out = "B1.tif"
  B2_out = "B2.tif"
  B3_out = "B3.tif"

#Copy raster and map algebra ??? makes new rasters of Red and NearIR
  arcpy.CopyRaster_management(B1, B1_out,"","0","0","","","8_BIT_UNSIGNED")
  print "Copied Band1 band as raster"
  arcpy.CopyRaster_management(B2, B2_out,"","0","0","","","8_BIT_UNSIGNED")
  print "Copied Band2 band as raster"
  arcpy.CopyRaster_management(B3, B3_out,"","0","0","","","8_BIT_UNSIGNED")
  print "Copied Band3 band as raster"

 
#Create Numerator and Denominator rasters as variables and NDVI output (note that arcpy.sa.Float returns a floating point raster)

  Num = arcpy.sa.Int(Raster(B1_out) + Raster(B2_out) + Raster(B3_out))
  Avg_band = arcpy.sa.Divide(Num, 3)
  print "Averaging"

#Saving output to result output you specified above.   Avg_band.save(result)
  print "Successful"
  print "Image number ", glob.glob("I:/alb_aerials/1957/*.jpg").index(line)
  print "Total elapsed time (hours): ", time.clock()/3600[/INDENT]

[ColinMurray]

Hello
I've been wondering about this for a while. To make a colour image greyscale without only saving 1 band (which I find reduces image quality).

Thanks. It looks like applying the Greyscale raster function and then exporting as a new raster will work.

For batch processing I'm going to try:
Using the edit raster function tool in a batch process for all the rasters.
Then perform another batch process with the export tool.