diff --git a/tools/points2binaryimage/test-data/output1.original.tif b/tools/points2binaryimage/test-data/output1.original.tif deleted file mode 100644 index 99a69a37..00000000 Binary files a/tools/points2binaryimage/test-data/output1.original.tif and /dev/null differ diff --git a/tools/points2labelimage/creators.xml b/tools/points2labelimage/creators.xml new file mode 120000 index 00000000..5d2b71e0 --- /dev/null +++ b/tools/points2labelimage/creators.xml @@ -0,0 +1 @@ +../../macros/creators.xml \ No newline at end of file diff --git a/tools/points2labelimage/points2label.py b/tools/points2labelimage/points2label.py index 7f701377..0d309a28 100644 --- a/tools/points2labelimage/points2label.py +++ b/tools/points2labelimage/points2label.py @@ -1,47 +1,125 @@ import argparse -import sys +import os import warnings +import giatools.pandas import numpy as np import pandas as pd +import scipy.ndimage as ndi import skimage.io +import skimage.segmentation -def points2label(labels, shape, output_file=None, has_header=False, is_TSV=True): - labelimg = np.zeros([shape[0], shape[1]], dtype=np.int32) +def rasterize(point_file, out_file, shape, has_header=False, swap_xy=False, bg_value=0, fg_value=None): - if is_TSV: - if has_header: - df = pd.read_csv(labels, sep='\t', skiprows=1, header=None) - else: - df = pd.read_csv(labels, sep='\t', header=None) - else: + img = np.full(shape, dtype=np.uint16, fill_value=bg_value) + if os.path.exists(point_file) and os.path.getsize(point_file) > 0: + + # Read the tabular file with information from the header if has_header: - df = pd.read_csv(labels, skiprows=1, header=None) + df = pd.read_csv(point_file, delimiter='\t') + + pos_x_column = giatools.pandas.find_column(df, ['pos_x', 'POS_X']) + pos_y_column = giatools.pandas.find_column(df, ['pos_y', 'POS_Y']) + pos_x_list = df[pos_x_column].round().astype(int) + pos_y_list = df[pos_y_column].round().astype(int) + assert len(pos_x_list) == len(pos_y_list) + + try: + radius_column = giatools.pandas.find_column(df, ['radius', 'RADIUS']) + radius_list = df[radius_column] + assert len(pos_x_list) == len(radius_list) + except KeyError: + radius_list = [0] * len(pos_x_list) + + try: + label_column = giatools.pandas.find_column(df, ['label', 'LABEL']) + label_list = df[label_column] + assert len(pos_x_list) == len(label_list) + except KeyError: + label_list = list(range(1, len(pos_x_list) + 1)) + + # Read the tabular file without header else: - df = pd.read_csv(labels, header=None) + df = pd.read_csv(point_file, header=None, delimiter='\t') + pos_x_list = df[0].round().astype(int) + pos_y_list = df[1].round().astype(int) + assert len(pos_x_list) == len(pos_y_list) + radius_list = [0] * len(pos_x_list) + label_list = list(range(1, len(pos_x_list) + 1)) + + # Optionally swap the coordinates + if swap_xy: + pos_x_list, pos_y_list = pos_y_list, pos_x_list - for i in range(0, len(df)): - a_row = df.iloc[i] - labelimg[a_row[0], a_row[1]] = i + 1 + # Perform the rasterization + for y, x, radius, label in zip(pos_y_list, pos_x_list, radius_list, label_list): + if fg_value is not None: + label = fg_value + + if y < 0 or x < 0 or y >= shape[0] or x >= shape[1]: + raise IndexError(f'The point x={x}, y={y} exceeds the bounds of the image (width: {shape[1]}, height: {shape[0]})') + + # Rasterize circle and distribute overlapping image area + if radius > 0: + mask = np.ones(shape, dtype=bool) + mask[y, x] = False + mask = (ndi.distance_transform_edt(mask) <= radius) + + # Compute the overlap (pretend there is none if the rasterization is binary) + if fg_value is None: + overlap = np.logical_and(img > 0, mask) + else: + overlap = np.zeros(shape, dtype=bool) + + # Rasterize the part of the circle which is disjoint from other foreground. + # + # In the current implementation, the result depends on the order of the rasterized circles if somewhere + # more than two circles overlap. This is probably negligable for most applications. To achieve results + # that are invariant to the order, first all circles would need to be rasterized independently, and + # then blended together. This, however, would either strongly increase the memory consumption, or + # require a more complex implementation which exploits the sparsity of the rasterized masks. + # + disjoint_mask = np.logical_xor(mask, overlap) + if disjoint_mask.any(): + img[disjoint_mask] = label + + # Distribute the remaining part of the circle + if overlap.any(): + dist = ndi.distance_transform_edt(overlap) + foreground = (img > 0) + img[overlap] = 0 + img = skimage.segmentation.watershed(dist, img, mask=foreground) + + # Rasterize point (there is no overlapping area to be distributed) + else: + img[y, x] = label - if output_file is not None: - with warnings.catch_warnings(): - warnings.simplefilter("ignore") - skimage.io.imsave(output_file, labelimg, plugin='tifffile') else: - return labelimg + raise Exception("{} is empty or does not exist.".format(point_file)) # appropriate built-in error? + + with warnings.catch_warnings(): + warnings.simplefilter("ignore") + skimage.io.imsave(out_file, img, plugin='tifffile') # otherwise we get problems with the .dat extension if __name__ == "__main__": parser = argparse.ArgumentParser() - parser.add_argument('label_file', type=argparse.FileType('r'), default=sys.stdin, help='label file') - parser.add_argument('out_file', type=argparse.FileType('w'), default=sys.stdin, help='out file') - parser.add_argument('org_file', type=argparse.FileType('r'), default=sys.stdin, help='input original file') - parser.add_argument('--has_header', dest='has_header', type=bool, default=False, help='label file has header') - parser.add_argument('--is_tsv', dest='is_tsv', type=bool, default=True, help='label file is TSV') - args = parser.parse_args() + parser.add_argument('point_file', type=argparse.FileType('r'), help='point file') + parser.add_argument('out_file', type=str, help='out file (TIFF)') + parser.add_argument('shapex', type=int, help='shapex') + parser.add_argument('shapey', type=int, help='shapey') + parser.add_argument('--has_header', dest='has_header', default=False, help='set True if point file has header') + parser.add_argument('--swap_xy', dest='swap_xy', default=False, help='Swap X and Y coordinates') + parser.add_argument('--binary', dest='binary', default=False, help='Produce binary image') - original_shape = skimage.io.imread(args.org_file.name, plugin='tifffile').shape + args = parser.parse_args() - points2label(args.label_file.name, original_shape, args.out_file.name, args.has_header, args.is_tsv) + rasterize( + args.point_file.name, + args.out_file, + (args.shapey, args.shapex), + has_header=args.has_header, + swap_xy=args.swap_xy, + fg_value=0xffff if args.binary else None, + ) diff --git a/tools/points2labelimage/points2label.xml b/tools/points2labelimage/points2label.xml index 27e0ce39..03a742e2 100644 --- a/tools/points2labelimage/points2label.xml +++ b/tools/points2labelimage/points2label.xml @@ -1,41 +1,121 @@ - + + + creators.xml + tests.xml + 0.4 + 0 + + + + operation_3443 galaxy_image_analysis - - numpy - scikit-image - pandas - tifffile - - - - - - - - + + scikit-image + numpy + pandas + tifffile + giatools + + + + + + + + + - + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + - - - + + + + + + + - **What it does** - This tool converts points to a label image. + **Converts a tabular list of points to a label map by rasterizing the point coordinates.** + + The created image is a single-channel image with 16 bits per pixel (unsigned integer). The points are + rasterized with unique labels, or the value 65535 (white) for binary image output. Pixels not corresponding to + any points in the tabular file are assigned the value 0 (black). + + The tabular list of points can either be header-less. In this case, the first and second columns are expected + to be the X and Y coordinates, respectively. Otherwise, if a header is present, it is searched for the + following column names: + + - ``pos_x`` or ``POS_X``: This column corresponds to the X coordinates. + - ``pos_y`` or ``POS_Y``: This column corresponds to the Y coordinates. + - If a ``radius`` or ``RADIUS`` column is present, then the points will be rasterized as circles of the + corresponding radii. + - If a ``label`` or ``LABEL`` column is present, then the corresponding labels will be used for rasterization + (unless "Produce binary image" is activated). Different points are allowed to use the same label. + 10.1016/j.jbiotec.2017.07.019 diff --git a/tools/points2labelimage/test-data/galaxyIcon_noText.tif b/tools/points2labelimage/test-data/galaxyIcon_noText.tif deleted file mode 100644 index 3dbc2efc..00000000 Binary files a/tools/points2labelimage/test-data/galaxyIcon_noText.tif and /dev/null differ diff --git a/tools/points2labelimage/test-data/input1.tsv b/tools/points2labelimage/test-data/input1.tsv new file mode 100644 index 00000000..ea2f865c --- /dev/null +++ b/tools/points2labelimage/test-data/input1.tsv @@ -0,0 +1,9 @@ +11.7555970149 10.4048507463 +15 14 +19 2 +5 4 +5 5 +5 6 +5 7 +5 8 +5 9 diff --git a/tools/points2labelimage/test-data/input2.tsv b/tools/points2labelimage/test-data/input2.tsv new file mode 100644 index 00000000..47efffa6 --- /dev/null +++ b/tools/points2labelimage/test-data/input2.tsv @@ -0,0 +1,39 @@ +frame pos_x pos_y scale radius intensity +1 85 32 1.33 3.77 18807.73 +1 190 25 1.78 5.03 24581.44 +1 137 26 1.44 4.09 19037.59 +1 63 42 1.44 4.09 22390.80 +1 107 44 1.33 3.77 23429.96 +1 61 27 1.56 4.40 18052.18 +1 158 39 1.44 4.09 18377.02 +1 190 14 1.33 3.77 18548.86 +1 182 33 1.78 5.03 26467.79 +1 39 39 1.44 4.09 14782.43 +1 169 26 1.33 3.77 14203.41 +1 61 54 1.33 3.77 23248.06 +1 95 52 1.33 3.77 21480.71 +1 23 60 1.89 5.34 25203.43 +1 84 24 1.56 4.40 16630.57 +1 121 47 1.67 4.71 15459.11 +1 66 49 1.11 3.14 23858.07 +1 115 36 2.00 5.66 16389.10 +1 55 51 1.33 3.77 23548.90 +1 130 72 1.67 4.71 15769.02 +1 117 23 1.33 3.77 16763.14 +1 45 52 1.56 4.40 22877.61 +1 36 71 1.56 4.40 20780.96 +1 78 17 1.33 3.77 16844.51 +1 101 38 1.56 4.40 21376.59 +1 147 31 1.78 5.03 16597.14 +1 163 55 2.00 5.66 18301.54 +1 164 23 1.33 3.77 17073.82 +1 150 24 1.56 4.40 15440.02 +1 151 67 1.78 5.03 18419.96 +1 26 53 2.00 5.66 20586.01 +1 79 62 1.33 3.77 15232.88 +1 69 17 1.11 3.14 15601.83 +1 83 52 1.33 3.77 18315.00 +1 16 54 2.00 5.66 22140.66 +1 166 61 1.78 5.03 18488.78 +1 163 43 1.44 4.09 16925.49 +1 130 53 1.78 5.03 15101.96 diff --git a/tools/points2labelimage/test-data/input3.tsv b/tools/points2labelimage/test-data/input3.tsv new file mode 100644 index 00000000..0193ae97 --- /dev/null +++ b/tools/points2labelimage/test-data/input3.tsv @@ -0,0 +1,4 @@ +pos_x pos_y radius label +20 20 20 1 +50 50 40 1 +150 50 40 2 diff --git a/tools/points2labelimage/test-data/out.tif b/tools/points2labelimage/test-data/out.tif deleted file mode 100644 index 7b73ed33..00000000 Binary files a/tools/points2labelimage/test-data/out.tif and /dev/null differ diff --git a/tools/points2labelimage/test-data/output1_binary.tif b/tools/points2labelimage/test-data/output1_binary.tif new file mode 100644 index 00000000..4a946253 Binary files /dev/null and b/tools/points2labelimage/test-data/output1_binary.tif differ diff --git a/tools/points2labelimage/test-data/output2.tif b/tools/points2labelimage/test-data/output2.tif new file mode 100644 index 00000000..3eb6a0df Binary files /dev/null and b/tools/points2labelimage/test-data/output2.tif differ diff --git a/tools/points2labelimage/test-data/output2_binary.tif b/tools/points2labelimage/test-data/output2_binary.tif new file mode 100644 index 00000000..922cd0d5 Binary files /dev/null and b/tools/points2labelimage/test-data/output2_binary.tif differ diff --git a/tools/points2labelimage/test-data/output3.tif b/tools/points2labelimage/test-data/output3.tif new file mode 100644 index 00000000..52e17335 Binary files /dev/null and b/tools/points2labelimage/test-data/output3.tif differ diff --git a/tools/points2labelimage/test-data/output3_binary.tif b/tools/points2labelimage/test-data/output3_binary.tif new file mode 100644 index 00000000..baa3b170 Binary files /dev/null and b/tools/points2labelimage/test-data/output3_binary.tif differ diff --git a/tools/points2labelimage/test-data/points.tsv b/tools/points2labelimage/test-data/points.tsv deleted file mode 100644 index 1949b6f8..00000000 --- a/tools/points2labelimage/test-data/points.tsv +++ /dev/null @@ -1,7 +0,0 @@ -5 3 -10 2 -8 4 -15 15 -8 5 -8 6 -8 7 diff --git a/tools/points2labelimage/tests.xml b/tools/points2labelimage/tests.xml new file mode 120000 index 00000000..e20d710a --- /dev/null +++ b/tools/points2labelimage/tests.xml @@ -0,0 +1 @@ +../../macros/tests.xml \ No newline at end of file