view imagej2_noise_jython_script.py @ 0:dd20ee3092e3 draft

"planemo upload for repository https://github.com/bgruening/galaxytools/tree/master/tools/image_processing/imagej2 commit b08f0e6d1546caaf627b21f8c94044285d5d5b9c-dirty"
author imgteam
date Tue, 17 Sep 2019 17:04:21 -0400
parents
children 53fb6f4afcc8
line wrap: on
line source

import sys
from ij import IJ
from ij import ImagePlus
import jython_utils

# Fiji Jython interpreter implements Python 2.5 which does not
# provide support for argparse.
error_log = sys.argv[ -19 ]
input = sys.argv[ -18 ]
image_datatype = sys.argv[ -17 ]
noise = sys.argv[ -16 ]
standard_deviation = sys.argv[ -15 ]
radius = sys.argv[ -14 ]
threshold = sys.argv[ -13 ]
which_outliers = sys.argv[ -12 ]
randomj = sys.argv[ -11 ]
trials = sys.argv[ -10 ]
probability = sys.argv[ -9 ]
# Note the spelling - so things don't get confused due to Python lambda function.
lammbda = sys.argv[ -8 ]
order = sys.argv[ -7 ]
mean = sys.argv[ -6 ]
sigma = sys.argv[ -5 ]
min = sys.argv[ -4 ]
max = sys.argv[ -3 ]
insertion = sys.argv[ -2 ]
tmp_output_path = sys.argv[ -1 ]

error = False

# Open the input image file.
image_plus = IJ.openImage( input )
bit_depth = image_plus.getBitDepth()
image_type = image_plus.getType()
# Create an ImagePlus object for the image.
image_plus_copy = image_plus.duplicate()
# Make a copy of the image.
image_processor_copy = image_plus_copy.getProcessor()

# Perform the analysis on the ImagePlus object.
if noise == 'add_noise':
    IJ.run( image_plus_copy, "Add Noise", "" )
elif noise == 'add_specified_noise':
    IJ.run( image_plus_copy, "Add Specified Noise", "standard=&standard_deviation" )
elif noise == 'salt_and_pepper':
    IJ.run( image_plus_copy, "Salt and Pepper", "" )
elif noise == 'despeckle':
    IJ.run( image_plus_copy, "Despeckle", "" )
elif noise == 'remove_outliers':
    IJ.run( image_plus_copy, "Remove Outliers", "radius=&radius threshold=&threshold which=&which_outliers" )
elif noise == 'remove_nans':
    if bit_depth == 32:
        IJ.run( image_plus_copy, "Remove NaNs", "" )
    else:
        # When Galaxy metadata for images is enhanced to include information like this,
        # we'll be able to write tool validators rather than having to stop the job in
        # an error state.
        msg = "Remove NaNs requires a 32-bit image, the selected image is %d-bit" % bit_depth
        jython_utils.handle_error( error_log, msg )
        error = True
elif noise == 'rof_denoise':
    if image_type == ImagePlus.GRAY32:
        IJ.run( image_plus_copy, "ROF Denoise", "" )
    else:
        msg = "ROF Denoise requires an image of type 32-bit grayscale, the selected image is %d-bit" % ( bit_depth )
        jython_utils.handle_error( error_log, msg )
        error = True
elif noise == 'randomj':
    if randomj == 'randomj_binomial':
        IJ.run( image_plus_copy, "RandomJ Binomial", "trials=&trials probability=&probability insertion=&insertion" )
    elif randomj == 'randomj_exponential':
        IJ.run( image_plus_copy, "RandomJ Exponential", "lambda=&lammbda insertion=&insertion" )
    elif randomj == 'randomj_gamma':
        IJ.run( image_plus_copy, "RandomJ Gamma", "order=&order insertion=&insertion" )
    elif randomj == 'randomj_gaussian':
        IJ.run( image_plus_copy, "RandomJ Gaussian", "mean=&mean sigma=&sigma insertion=&insertion" )
    elif randomj == 'randomj_poisson':
        IJ.run( image_plus_copy, "RandomJ Poisson", "mean=&mean insertion=&insertion" )
    elif randomj == 'randomj_uniform':
        IJ.run( image_plus_copy, "RandomJ Uniform", "min=&min max=&max insertion=&insertion" )

if not error:
    # Save the ImagePlus object as a new image.
    IJ.saveAs( image_plus_copy, image_datatype, tmp_output_path )