From a7220a243e2a7851f45d94e11a5b552256d596cf Mon Sep 17 00:00:00 2001
From: monadal <morgane.nadal@inria.fr>
Date: Wed, 11 Mar 2020 10:29:54 +0100
Subject: [PATCH] fct for intensity normalization and find voxel dimension in
micron + refractoring
---
brick/processing/input.py | 39 +++++++++++++++++++++++++++++++++------
nutrimorph.py | 4 ++++
2 files changed, 37 insertions(+), 6 deletions(-)
diff --git a/brick/processing/input.py b/brick/processing/input.py
index a9a1a11..56a081f 100644
--- a/brick/processing/input.py
+++ b/brick/processing/input.py
@@ -32,9 +32,11 @@
from brick.general.type import array_t
import numpy as np_
import sys as sy_
+from PIL import Image
+from PIL.ExifTags import TAGS
-def ImageVerification(image, channel):
+def ImageVerification(image: array_t, channel: str) -> array_t:
if image.ndim == 3:
print('Your image has only one color channel.')
if channel == 'RGB':
@@ -59,6 +61,7 @@ def ImageVerification(image, channel):
print('The image dimensions are not correct:', image.ndim, ', instead of 3 or 4.')
sy_.exit(0)
+
# if channel != 'RGB' and image.ndim == 4:
# print('The image has multiple color channels. The channel', channel, 'was specified in the parameters.')
# for idx, color in enumerate('RGB'):
@@ -70,12 +73,13 @@ def ImageVerification(image, channel):
def IntensityNormalizedImage(image: array_t) -> array_t:
#
- print('Relative Intensity Normalization between 0 and 1 (Not a standardization). Need to reevaluate the parameters !!!')
+ print(
+ 'Relative Intensity Normalization between 0 and 1 (Not a standardization). Need to reevaluate the parameters !!!')
value_max = image.astype(np_.float32).max()
value_min = image.astype(np_.float32).min()
- result = (image.astype(np_.float32) - value_min)/(value_max - value_min)
+ result = (image.astype(np_.float32) - value_min) / (value_max - value_min)
return result
@@ -92,9 +96,32 @@ def IntensityNormalizedImage(image: array_t) -> array_t:
#
# return result
-def MetricNormalizedImage(image: array_t) -> array_t: # TODO
- #
- print('Image metric in converted to micrometers.')
+
+def VoxelDimensionInMicrons(data_path: str) -> array_t:
+ # TODO Verify if metadata are in the same format for all the images - if not, add a raise error ?
+
+ with Image.open(data_path) as img: # Find the voxels dimensions in micron in the metadata.
+ meta_dict = {TAGS.get(key, 'missing'): img.tag[key] for key in
+ img.tag} # Use the exif tags into the image metadata
+
+ # Decode the tags text
+ metadata = meta_dict['missing'].decode('utf8') # Decode the tags text
+ metadata = metadata.replace('\x00', '')
+
+ voxel_size = ['', '', ''] # Initialize the list of voxel size in str
+ for idaxe, axe in enumerate('XYZ'):
+ idvox = metadata.find('dblVoxel' + axe) # Get the index of the voxel size for the axe X, Y, Z
+ idvox += 15
+ while metadata[idvox] != '\n':
+ voxel_size[idaxe] += metadata[idvox]
+ idvox += 1
+ voxel_size = np_.asarray(list(map(float, voxel_size)))
+ voxel_size_microns = 1.0e6 * voxel_size
+ print('Voxel dimension in the image is [X Y Z] =', voxel_size_microns, 'in microns. WARNING this method highly '
+ 'depends on the format of the metadata.')
+
+ return voxel_size_microns
+
def DijkstraCosts(image: array_t, som_map: array_t, ext_map: array_t) -> array_t:
#
diff --git a/nutrimorph.py b/nutrimorph.py
index ca15e44..52d993c 100644
--- a/nutrimorph.py
+++ b/nutrimorph.py
@@ -54,6 +54,7 @@ import numpy as np_
import skimage.io as io_
import skimage.measure as ms_
from skimage.segmentation import relabel_sequential
+import exifread as xf_
print(sy_.argv, sy_.argv.__len__())
@@ -89,6 +90,9 @@ start_time = tm_.time()
# --- Images
+voxel_micron = in_.VoxelDimensionInMicrons(data_path)
+
+
image = io_.imread(data_path)
# Image size verification - simple version without user interface
--
GitLab