# HG changeset patch # User imgteam # Date 1549741654 18000 # Node ID 597b7ef44b05a69b7016ee0a26901bb11b02a52e planemo upload for repository https://github.com/BMCV/galaxy-image-analysis/tools/split_labelmaps/ commit c3f4b766f03770f094fda6bda0a5882c0ebd4581 diff -r 000000000000 -r 597b7ef44b05 split_labelmap.py --- /dev/null Thu Jan 01 00:00:00 1970 +0000 +++ b/split_labelmap.py Sat Feb 09 14:47:34 2019 -0500 @@ -0,0 +1,75 @@ +from imageio import imread as io_imread +from skimage.measure import regionprops +import numpy as np +#import matplotlib.pyplot as plt +import scipy +import skimage.io +import skimage.draw +from tifffile import imsave +import os +import argparse +import warnings + +# split_label_image takes a label image and outputs a similar file with the given name where the labeled +# parts of the image that touch (or overlap) are separated by at least 1 pixel (at most 2). + + +def split_labelmap(labelmap,outputfile): + + # Information from the label map. + label_img = io_imread(labelmap) + xtot, ytot = label_img.shape + props = regionprops(label_img) + N = len(props) + + # Creating the backgrounds. + background = np.zeros([xtot,ytot], 'uint8') + overlap = np.zeros([N,xtot,ytot],'uint8') + compstruct = scipy.ndimage.generate_binary_structure(2, 2) # Mask for image dilation. + + i = 0 + for cell in props: + cell_image = cell.image.astype('uint8') + #plt.imshow(cell_image) + + # Replace the background area corresponding to the bounding box with the image representing the cell. + background[int(cell.bbox[0]):int(cell.bbox[2]),int(cell.bbox[1]):int(cell.bbox[3])] += cell_image + overlap[i][int(cell.bbox[0]):int(cell.bbox[2]), int(cell.bbox[1]):int(cell.bbox[3])] = cell_image + + # In the overlap array, dilate the cell in all directions. + overlap[i] = scipy.ndimage.binary_dilation( + overlap[i], structure=compstruct).astype(overlap[i].dtype) + + i += 1 + + if len(props) > 1: + # Sum together the overlap. + total_overlap = sum(overlap) + + # Wherever the overlap is greater than 1 replace that point with zero in the final image. + for x in range(xtot): + for y in range(ytot): + if total_overlap[x,y] > 1: + background[x,y] = 0 + + # Force the image into 8-bit. + result = skimage.util.img_as_ubyte(background) + + # Save image + with warnings.catch_warnings(): + warnings.simplefilter("ignore") + skimage.io.imsave(outputfile, result, plugin="tifffile") + + return None + +# To run from command line. +if __name__ == "__main__": + parser = argparse.ArgumentParser() + parser.add_argument('labelmap', + help='Label map image.') + parser.add_argument('outputfile', + help='Output file. Without extension (although it corrects if you ' + 'add it; will always return a .tif') + + args = parser.parse_args() + split_labelmap(args.labelmap, args.outputfile) diff -r 000000000000 -r 597b7ef44b05 split_labelmap.xml --- /dev/null Thu Jan 01 00:00:00 1970 +0000 +++ b/split_labelmap.xml Sat Feb 09 14:47:34 2019 -0500 @@ -0,0 +1,34 @@ + + + + scikit-image + numpy + tifffile + + + + + + + + + + + + + + + + + + **What it does** + + Takes a labeled image and outputs a similar file where the labeled parts + of the image that touch (or overlap) are separated by at least 1 pixel (at most 2). + + + 10.1016/j.jbiotec.2017.07.019 + + diff -r 000000000000 -r 597b7ef44b05 test-data/out.tif Binary file test-data/out.tif has changed diff -r 000000000000 -r 597b7ef44b05 test-data/sample_seg.tif Binary file test-data/sample_seg.tif has changed