Skip to content

Commit

Permalink
Merge pull request #742 from danforthcenter/visualize_hyper_histogram
Browse files Browse the repository at this point in the history
Add a function that plots histograms for hyperspectral images
  • Loading branch information
nfahlgren authored Jul 23, 2021
2 parents ee89b2b + 2b75fdd commit 40497c5
Show file tree
Hide file tree
Showing 12 changed files with 342 additions and 16 deletions.
Loading
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Loading
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Loading
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Loading
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Loading
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Loading
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
5 changes: 5 additions & 0 deletions docs/updating.md
Original file line number Diff line number Diff line change
Expand Up @@ -961,6 +961,11 @@ pages for more details on the input and output variable types.
* post v3.5: fig_hist = **plantcv.visualize.histogram**(*gray_img, mask=None, bins=256, color='red', title=None*)
* post v3.12: fig_hist, hist_data = **plantcv.visualize.histogram**(*img, mask=None, bins=100, lower_bound=None, upper_bound=None, title=None, hist_data=False*)

#### plantcv.visualize.hyper_histogram

* pre v3.13: NA
* post v3.13: fig_hist = **plantcv.visualize.hyper_histogram**(*hsi, mask=None, bins=100, lower_bound=None, upper_bound=None, title=None, wvlengths=[480, 550, 650]*)

#### plantcv.visualize.obj_size_ecdf

* pre v3.13: NA
Expand Down
78 changes: 78 additions & 0 deletions docs/visualize_hyper_histogram.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,78 @@
## Plot Histograms for Hyperspectral Images

This is a plotting method that is used to examine the distribution of signal within an hyperspectral image for
specific wavelengths. This function is an application of the [histogram](visualize_histogram.md) function.

**plantcv.visualize.hyper_histogram**(*hsi, mask=None, bins=100, lower_bound=None, upper_bound=None, title=None, wvlengths=[480, 550, 670]*)

**returns** fig_hist

- **Parameters:**
- hsi - [Spectral_data](Spectral_data.md)
- mask - Optional binary mask made from selected contours.
- bins - Number of class to divide spectrum into (default bins=100).
- lower_bound - Lower bound of range to be shown in the histogram (default lower_range=None).
- upper_bound - Upper bound of range to be shown in the histogram.
- title - The title for the histogram (default title=None)
- wvlengths - Interested wavelengths (in nanometers) to show histogram of (default wvlengths=[480, 550, 670])

- **Context:**
- Examine the distribution of the signal, this can help select a value for calling the binary thresholding function.
- If the provided wavelength is in the range of visible spectrum, the color corresponds to the wavelength would be used as the histogram color

- **Example use:**
<!---
- [Use In NIR Tutorial](nir_tutorial.md)
--->

**Hyperspectral image**

![Screenshot](img/documentation_images/hyper_histogram/hyper.png)

**Mask**

![Screenshot](img/documentation_images/hyper_histogram/mask.png)

```python

from plantcv import plantcv as pcv

pcv.params.debug = "plot"

# Showing histograms for 3 default wavelengths
hist_figure1 = pcv.visualize.hyper_histogram(hsi=hsi, mask=mask)

# Showing the histogram for a single wavelength (700nm)
hist_figure2 = pcv.visualize.hyper_histogram(hsi=hsi, mask=mask, wvlengths=[700])

# Showing the histogram for two wavelengths
hist_figure3 = pcv.visualize.hyper_histogram(hsi=hsi, mask=mask, wvlengths=[380, 970])

# Showing the histogram for multiple wavelengths
hist_figure4 = pcv.visualize.hyper_histogram(hsi=hsi, mask=mask,
wvlengths=[379, 409, 439, 469, 499, 529, 559, 568,
619, 649, 679, 709, 739, 769, 799,
829, 859, 889,
919, 949, 979, 1000])

```

**Histogram of signal intensity**

![hist_default](img/documentation_images/hyper_histogram/hist_default_bands.png)

![hist_single](img/documentation_images/hyper_histogram/hist_single_band.png)

![hist_two](img/documentation_images/hyper_histogram/hist_two_bands.png)

![hist_multiple](img/documentation_images/hyper_histogram/hist_multiple_bands.png)

To plot all wavelengths (not recommended for hyperspectral images supporting a large number of bands):

```python
# Plot not shown
hist_all_wv = pcv.visualize.hyper_histogram(hsi=hsi, mask=mask, wvlengths=list(hsi.wavelength_dict.keys()))

```

**Source Code:** [Here](https://github.com/danforthcenter/plantcv/blob/master/plantcv/plantcv/visualize/hyper_histogram.py)
1 change: 1 addition & 0 deletions mkdocs.yml
Original file line number Diff line number Diff line change
Expand Up @@ -164,6 +164,7 @@ nav:
- 'Colorize Masks': visualize_colorize_masks.md
- 'Colorspaces': visualize_colorspace.md
- 'Histogram': visualize_histogram.md
- 'Hyperspectral Histogram': visualize_hyper_histogram.md
- 'Overlay Images': visualize_ovelay_two_imgs.md
- 'Object Size ECDF': visualize_obj_size_ecdf.md
- 'Object Sizes': visualize_obj_sizes.md
Expand Down
4 changes: 2 additions & 2 deletions plantcv/plantcv/visualize/__init__.py
Original file line number Diff line number Diff line change
Expand Up @@ -8,7 +8,7 @@
from plantcv.plantcv.visualize.colorize_label_img import colorize_label_img
from plantcv.plantcv.visualize.obj_sizes import obj_sizes
from plantcv.plantcv.visualize.obj_size_ecdf import obj_size_ecdf

from plantcv.plantcv.visualize.hyper_histogram import hyper_histogram

__all__ = ["pseudocolor", "colorize_masks", "histogram", "clustered_contours", "colorspaces", "auto_threshold_methods",
"overlay_two_imgs", "colorize_label_img", "obj_size_ecdf", "obj_sizes"]
"overlay_two_imgs", "colorize_label_img", "obj_size_ecdf", "obj_sizes", "hyper_histogram"]
196 changes: 196 additions & 0 deletions plantcv/plantcv/visualize/hyper_histogram.py
Original file line number Diff line number Diff line change
@@ -0,0 +1,196 @@
# Help visualize histograms for hyperspectral images

import os
import numpy as np
import pandas as pd
from plantcv.plantcv._debug import _debug
from plantcv.plantcv import fatal_error, params, color_palette
from plotnine import ggplot, aes, geom_line, scale_color_manual, theme_classic, labels
from plantcv.plantcv.visualize import histogram
from plantcv.plantcv.hyperspectral import _find_closest
import math


def _get_color_dict_uv():
"""Create a color dictionary for UV wavelengths."""
params.color_scale = "cool_r"
uv_wavelengths = np.arange(290, 444)
uv_colors_ = color_palette(num=256)
uv_colors_ = uv_colors_[0:len(uv_wavelengths)]
uv_colors_ = [tuple([xi / 255 for xi in x]) for x in uv_colors_[::-1]]
uv_colors = {}
for i, wv in enumerate(uv_wavelengths):
uv_colors[wv] = uv_colors_[i]
return uv_colors


def _get_color_dict_vis():
"""Create a color dictionary for VIS wavelengths."""
params.color_scale = "turbo"
vis_wavelengths = np.arange(445, 701)
vis_colors_ = color_palette(num=256)
vis_colors_ = [tuple([xi / 255 for xi in x]) for x in vis_colors_]
vis_colors = {}
for i, wv in enumerate(vis_wavelengths):
vis_colors[wv] = vis_colors_[i]
return vis_colors


def _get_color_dict_nir():
"""Create a color dictionary for infrared wavelengths."""
params.color_scale = "inferno"
nir_wavelengths = np.arange(701, 1725)
# nir_wavelengths = [_round_to_multiple(x, multiple=4, min_wv=701, max_wv=1725) for x in nir_wavelengths_]
nir_colors_ = color_palette(num=256)
nir_colors_ = [tuple([xi / 255 for xi in nir_colors_[math.floor(idx / 4)]]) for (idx, _) in
enumerate(nir_wavelengths)]
nir_colors = {}
for i, wv in enumerate(nir_wavelengths):
nir_colors[wv] = nir_colors_[i]
return nir_colors


def _rgb_to_webcode(rgb_values):
"""
Convert RGB values to webcodes.
Inputs:
rgb_value: a tuple of RGB values (0~1, float)
Returns:
webcode: a webcode string encoding the RGB values
"""
webcode = "#"
for value in rgb_values:
code_ = hex(int(value*255)).replace('0x', '')
code = code_.upper() if len(code_) > 1 else '0{}'.format(code_.upper())
webcode += code
return webcode


def hyper_histogram(hsi, mask=None, bins=100, lower_bound=None, upper_bound=None,
title=None, wvlengths=[480, 550, 650]):
"""
Plot a histograms of selected wavelengths from a hyperspectral image.
This function calculates the histogram of selected wavelengths hyperspectral images
The color of the histograms are based on the wavelength if the wavelength is in the range of visible spectrum;
otherwise, random colors are assigned
Inputs:
hsi = Hyperspectral data instance
mask = binary mask, if provided, calculate histogram from masked area only (default=None)
bins = divide the data into n evenly spaced bins (default=100)
lower_bound = the lower bound of the bins (x-axis min value) (default=None)
upper_bound = the upper bound of the bins (x-axis max value) (default=None)
title = a custom title for the plot (default=None)
wvlengths = (optional) list of wavelengths to show histograms (default = [480,550,670], i.e. only show
histograms for blue, green, and red bands)
Returns:
fig_hist = histogram figure
:param hsi: plantcv.plantcv.classes.Spectral_data
:param mask: numpy.ndarray
:param bins: int
:param lower_bound: None, int, float
:param upper_bound: None, int, float
:param title: None, str
:param wvlengths: list
:return fig_hist: plotnine.ggplot.ggplot
"""

# Always sort desired wavelengths
wvlengths.sort()

# Available wavelengths of the spectral data
wl_keys = hsi.wavelength_dict.keys()
wls = np.array([float(i) for i in wl_keys])
wls.sort()

# Spectral resolution of the spectral data
diffs = [wls[i] - wls[i - 1] for i in range(1, len(wls))]
spc_res = sum(diffs) / len(diffs)

# Check if the distance is greater than 2x the spectral resolution
# If the distance > 2x resolution, it is considered being out of available ranges
checks = []
for i in range(0, len(wvlengths)):
checks.append(hsi.min_wavelength - wvlengths[i] > 2 * spc_res)
checks.append(wvlengths[i] - hsi.max_wavelength > 2 * spc_res)
if np.any(checks):
fatal_error(f"At least one band is too far from the available wavelength range: "
f"({hsi.min_wavelength},{hsi.max_wavelength})!")

# Find indices of bands whose wavelengths are closest to desired ones
match_ids = [_find_closest(wls, wv) for wv in wvlengths]
match_wls = [round(wls[i]) for i in match_ids]

# prepare color dictionary(ies)
color_dict = {}
if any(x < 290 for x in match_wls):
# under uv
params.color_scale = "cool_r"
color_ = color_palette(num=256)[-154]
under_uv_colors_ = {}
for i, wv in enumerate([x for x in match_wls if x < 290]):
under_uv_colors_[wv] = color_
color_dict = {**color_dict, **under_uv_colors_}
if any(290 <= x < 445 for x in match_wls):
uv_colors = _get_color_dict_uv()
color_dict = {**color_dict, **uv_colors}
if any(445 <= x < 701 for x in match_wls):
vis_colors = _get_color_dict_vis()
color_dict = {**color_dict, **vis_colors}
if any(701 <= x < 1701 for x in match_wls):
nir_colors = _get_color_dict_nir()
color_dict = {**color_dict, **nir_colors}
if any(x >= 1701 for x in match_wls):
# above nir
params.color_scale = "inferno"
color_ = color_palette(num=256)[-1]
above_uv_colors_ = {}
for i, wv in enumerate([x for x in match_wls if x >= 1701]):
above_uv_colors_[wv] = color_
color_dict = {**color_dict, **above_uv_colors_}

colors_rgb = [color_dict[wv] for wv in match_wls]
colors_hex = [_rgb_to_webcode(x) for x in colors_rgb]

array_data = hsi.array_data

# List of wavelengths recorded created from parsing the header file will be string, make list of floats
histograms = dict()
hist_dataset = pd.DataFrame(columns=['reflectance'])
debug = params.debug
params.debug = None

# Create a dataframe for all histogram related information
# (using the "histogram" function in "visualization" subpackage)
for i_wv, (wv, color) in enumerate(zip(wvlengths, colors_hex)):
idx = match_ids[i_wv]
_, hist_data = histogram(array_data[:, :, idx], mask=mask, bins=bins, lower_bound=lower_bound,
upper_bound=upper_bound, title=title, hist_data=True)
histograms[wv] = {"label": wv, "graph_color": color, "reflectance": hist_data['pixel intensity'].tolist(),
"hist": hist_data['proportion of pixels (%)'].tolist()}
if i_wv == 0:
hist_dataset['reflectance'] = hist_data['pixel intensity'].tolist()
hist_dataset[wv] = hist_data['proportion of pixels (%)'].tolist()

# Make the histogram figure using plotnine
df_hist = pd.melt(hist_dataset, id_vars=['reflectance'], value_vars=wvlengths,
var_name='Wavelength (' + hsi.wavelength_units + ')', value_name='proportion of pixels (%)')

fig_hist = (ggplot(df_hist, aes(x='reflectance', y='proportion of pixels (%)',
color='Wavelength (' + hsi.wavelength_units + ')'))
+ geom_line()
+ scale_color_manual(colors_hex, expand=(0, 0))
+ theme_classic()
)
if title is not None:
fig_hist = fig_hist + labels.ggtitle(title)

params.debug = debug
_debug(fig_hist, filename=os.path.join(params.debug_outdir, str(params.device) + '_histogram.png'))

return fig_hist
Loading

0 comments on commit 40497c5

Please sign in to comment.