ift.SetSuperpixelIGraphFeatures(igraph, feats, ift.WeightedL2NormPtr(),
0.2, 12.0)
suppixel = ift.SuperpixelSegmentationByRISF(igraph, opt, img)
B = ift.Circular(1.5)
return ift.RelabelRegions(suppixel, B)
if __name__ == '__main__':
parser = argparse.ArgumentParser(
prog='Image segmentation using optimum-path trees and seed selection.')
parser.add_argument('-i',
'--input-image',
type=check_io_img,
required=True)
parser.add_argument('-n', '--number-superpixels', type=int, required=True)
parser.add_argument('-o',
'--output-image',
type=check_aux_img,
default='output.pgm',
required=False)
args = parser.parse_args()
img = ift.ReadImageByExt(args.input_image)
n = args.number_superpixels
spxl = RISF(img, n)
ift.WriteImageByExt(spxl, args.output_image)
import pyift.pyift as ift
import sys
if __name__ == "__main__":
if len(sys.argv) != 3:
print(
sys.argv[0] +
" <imagem de entrada (jpg,png,pgm)> <imagem de saida (jpg,png,pgm)>"
)
else:
img = ift.ReadImageByExt(sys.argv[1])
A = ift.Rectangular(4, 4)
img = ift.MeanFilter(img, A)
A = ift.Rectangular(2, 2)
img = ift.Erode(img, A, None)
img = ift.Threshold(img, 0, 220, 255)
img = ift.Complement(img)
img = ift.Dilate(img, A, None)
ift.WriteImageByExt(img, sys.argv[2])
seeds_2_dir = sorted([
pjoin(scrib_2_dir, f) for f in os.listdir(scrib_2_dir)
if f.endswith('.txt')
])
A = ift.Circular(1.0)
if not os.path.exists(out_1_dir):
os.makedirs(out_1_dir)
if not os.path.exists(out_2_dir):
os.makedirs(out_2_dir)
for idx, path in enumerate(origs_dir):
img = ift.ReadImageByExt(path)
mimg = ift.ImageToMImage(img, ift.LABNorm_CSPACE)
seeds_1 = ift.ReadSeeds(img, seeds_1_dir[idx])
seeds_2 = ift.ReadSeeds(img, seeds_2_dir[idx])
print('\nSegmenting image {} ...'.format(idx + 1))
print('Scribbles 1')
print('Dynamic Tree')
segm = ift.DynTreeRoot(mimg, A, seeds_1, 1, 0.0, None, 0.0)
ift.WriteImageByExt(
ift.Mask(img, segm),
pjoin(out_1_dir, '{:02d}'.format(idx + 1) + '-dyntree.ppm'))
print('Dynamic Tree + Neighbor Dist')
default=0.5,
required=False,
help=
'Alpha needs to be tuned according to original image and saliency depth.'
)
parser.add_argument('-d', '--delta', type=int, default=0, required=False)
args = parser.parse_args()
if args.gamma < 0.0:
raise argparse.ArgumentTypeError('Gamma must be greater than 0.0.')
if args.alpha < 0.0 or args.alpha > 1.0:
raise argparse.ArgumentTypeError('Alpha must be between 0.0 and 1.0.')
orig = ift.ReadImageByExt(args.input_image)
mrk = ift.ReadImageByExt(args.markers_image)
seeds = ift.LabeledSetFromSeedImage(mrk, True)
if args.saliency_image != '':
objmap = ift.ReadImageByExt(args.saliency_image)
else:
objmap = None
mimg = ift.ImageToMImage(orig, ift.LABNorm_CSPACE)
A = ift.Circular(1.0)
if args.closest_root:
segm = ift.DynTreeClosestRoot(mimg, A, seeds, args.delta, args.gamma,
objmap, args.alpha)
else:
raise argparse.ArgumentTypeError('Auxiliary images must be a .pgm file.')
if __name__ == '__main__':
parser = argparse.ArgumentParser(prog='Object saliency using optimum-path forest and seed selection.')
parser.add_argument('-i', '--input-image', type=check_io_img, required=True)
parser.add_argument('-m', '--markers-image', type=check_aux_img, required=True)
parser.add_argument('-s', '--superpixel-image', type=check_aux_img, required=True)
parser.add_argument('-o', '--output-image', type=check_io_img, default='output.pgm', required=False)
parser.add_argument('-k', '--number-neighbors', type=int, default=24, required=False)
parser.add_argument('-t', '--threshold', type=float, default=0.8, required=False)
args = parser.parse_args()
img = ift.ReadImageByExt(args.input_image)
mrk = ift.ReadImageByExt(args.markers_image)
spxl = ift.RelabelImage(ift.ReadImageByExt(args.superpixel_image))
A = ift.Circular(1.5)
# seed selection
feat = ift.ExtractImageFeatures(img, spxl, A, False)
Z = ift.MImageToDataSet(feat, spxl)
ift.SetStatus(Z, ift.IFT_TRAIN)
graph = ift.CreateKnnGraph(Z, args.number_neighbors)
ift.UnsupTrain(graph, ift.NormalizedCutPtr())
cluster = ift.DataSetToLabelImage(Z, spxl, False, ift.IFT_GROUP)
seeds = ift.SelectSeedsForEnhancement(ift.LabeledSetFromSeedImage(mrk, True),
cluster, 1, args.threshold)