Skimage learning (2) -- RGB to grayscale, RGB to HSV, histogram matching

1、RGB Go gray

This example will have RGB The image of the channel is converted into an image with a single gray channel .

The value of each gray pixel is calculated as the corresponding red 、 Weighted sum of green and blue pixels ：

Y = 0.2125 R + 0.7154 G + 0.0721 B

CRT Phosphors use these weights , Because they are more representative of human beings to red than the same weight 、 The perception of green and blue .

import matplotlib.pyplot as plt

from skimage import data
from skimage.color import rgb2gray

original = data.astronaut()
grayscale = rgb2gray(original)
#fig It's a variable name ,fig Represents the drawing window (Figure);ax Represents the coordinate system on this drawing window (axis), Will generally continue to ax To operate 
# among figsize Used to set the size of the drawing ,a Is the width of the figure , b Is the height of the figure , In inches .
fig, axes = plt.subplots(1, 2, figsize=(8, 4))
ax = axes.ravel()

ax[0].imshow(original)
ax[0].set_title("Original")
ax[1].imshow(grayscale, cmap=plt.cm.gray)
ax[1].set_title("Grayscale")

fig.tight_layout()#ight_layout The subgraph parameters will be automatically adjusted , Make it fill the entire image area . This is an experimental feature , May not work in some cases . It just checks the axis labels 、 Scale label and title section .
plt.show()

2、RGB turn HSV

hsv Detailed explanation ：https://blog.csdn.net/bamboocan/article/details/70627137
This example shows how RGB To HSV( tonal , saturation , brightness ) Transformation can be used to facilitate the segmentation process .
Usually , The objects in the image have different colors ( tonal ) And brightness , So these features can be used to separate different areas of the image .
stay RGB In the middle , Hue and brightness are expressed as R、G、B Linear combination of channels , And they correspond to HSV Single channel of image ( Hue and value channels ).
A simple segmentation image , Then, threshold only HSV passageway .

For example, picture dyeing homogenization , We often use rgb2hsv To solve the problem , It can effectively protect the texture 、 Distribution and other characteristics .
Here we look at an example of chromosome color separation ：
Be careful , The example here is HED Color space ,HSV Space is the same effect .

from skimage import io
import matplotlib.pyplot as plt

import matplotlib.pyplot as plt

from skimage import data
from skimage.color import rgb2hsv
# First load RGB Image and extract Hue and Value passageway :
rgb_img = data.coffee()
hsv_img = rgb2hsv(rgb_img)
io.imshow(hsv_img)
plt.show()
hue_img = hsv_img[:, :, 0]# It is a way to process multidimensional data , It represents the first two dimensions , Take all of them 0 Index number .
value_img = hsv_img[:, :, 2]
value1_img = hsv_img[:, :, 1]

fig, (ax0, ax1, ax2,ax3) = plt.subplots(ncols=4, figsize=(8, 2))

ax0.imshow(rgb_img)
ax0.set_title("RGB image")
ax0.axis('off')
ax1.imshow(hue_img, cmap='hsv')
ax1.set_title("Hue channel")
ax1.axis('off')
ax2.imshow(value_img)
ax2.set_title("Value channel")
ax2.axis('off')
ax3.imshow(value1_img)
ax3.set_title("Value1 channel")
ax3.axis('off')

fig.tight_layout()
# stay Hue Set a threshold on the channel , Separate the cup from the background :
hue_threshold = 0.04# Why is this passage 0.04
binary_img = hue_img > hue_threshold

fig, (ax0, ax1) = plt.subplots(ncols=2, figsize=(8, 3))

''' plt.hist  Function is used to draw histogram . The function prototype ： plt.hist(x, bins=None)  Parameters  x  It's a one-dimensional array ,bins  It can be understood as the number of rectangles , The default is  10. '''
ax0.hist(hue_img.ravel(), 512)#512  Is the number of histogram bars   Set by yourself 
ax0.set_title("Histogram of the Hue channel with threshold")#Hue Channel histogram with threshold 
#matplotlib Library axiss Module Axes.axvline() The function is used to add a vertical line on the axis .
ax0.axvline(x=hue_threshold, color='r', linestyle='dashed', linewidth=2)
#matplotlib Library axiss Module Axes.set_xbound() Function to set x The upper and lower numerical boundaries of the axis .
ax0.set_xbound(0, 0.12)
ax1.imshow(binary_img)
ax1.set_title("Hue-thresholded image")
ax1.axis('off')

fig.tight_layout()
# stay Value Perform an additional threshold on the channel , Partially remove the shadow of the cup :
fig, ax0 = plt.subplots(figsize=(4, 3))

value_threshold = 0.10
binary_img = (hue_img > hue_threshold) | (value_img < value_threshold)

ax0.imshow(binary_img)
ax0.set_title("Hue and value thresholded image")
ax0.axis('off')

fig.tight_layout()
plt.show()

3、 Histogram matching

This example shows the characteristics of histogram matching . It manipulates the pixels of the input image , Match its histogram with the histogram of the reference image .
If the image has more than one channel , As long as the number of channels in the input image and the reference image is equal , Then match each channel independently .
Histogram matching can be used as a lightweight normalization in image processing , For example, feature matching , Especially in images from different sources or different conditions ( Namely light ) In the case of shooting .

import matplotlib.pyplot as plt

from skimage import data
from skimage import exposure
from skimage.exposure import match_histograms

reference = data.coffee()
image = data.chelsea()

matched = match_histograms(image, reference, channel_axis=-1)
''' '''
fig, (ax1, ax2, ax3) = plt.subplots(nrows=1, ncols=3, figsize=(8, 3),
                                    sharex=True, sharey=True)
for aa in (ax1, ax2, ax3):
    aa.set_axis_off()

ax1.imshow(image)
ax1.set_title('Source')
ax2.imshow(reference)
ax2.set_title('Reference')
ax3.imshow(matched)
ax3.set_title('Matched')

plt.tight_layout()
plt.show()
# To illustrate the effect of histogram matching , We draw each RGB Histogram and cumulative histogram of channels .
#  Obviously , The matching image of each channel has the same cumulative histogram as the reference image .
fig, axes = plt.subplots(nrows=3, ncols=3, figsize=(8, 8))

#enumerate This adds an index , At the same time, it can read the elements 
for i, img in enumerate((image, reference, matched)):
    for c, c_color in enumerate(('red', 'green', 'blue')):
        img_hist, bins = exposure.histogram(img[..., c], source_range='dtype')
        # The largest value of each group of data divided by the data is normalization . The maximum value after processing is 1, Other values are less than  1  Number of numbers .
        # In the histogram bin The meaning of ： To calculate the color histogram, we need to divide the color space into several small color intervals , That is, histogram bin, The color histogram is obtained by calculating the inner pixels of color in each cell ,bin The more , The stronger the resolution of histogram to color , But it increases the burden of computers . namely （ As shown in the figure above 10 Vertical bar area , Each vertical bar area is called a bin）
        axes[c, i].plot(bin, img_hist / img_hist.max())

        img_cdf, bins = exposure.cumulative_distribution(img[..., c])
        axes[c, i].plot(bins, img_cdf)
        axes[c, 0].set_ylabel(c_color)
        print(c,i)
axes[0, 0].set_title('Source')
axes[0, 1].set_title('Reference')
axes[0, 2].set_title('Matched')

plt.tight_layout()
plt.show()

annotation ：
1、 Returns the histogram of the image

cucim.skimage.exposure.histogram(image, nbins=256, source_range='image', normalize=False)

Returns the histogram of the image , And numpy.histogram Different , This function returns bin Center of , And will not recombine integer arrays .
For integer arrays , Each integer value has its own bin, This improves speed and intensity-resolution.
Calculate histogram on flattened image ： For color images , This function should be used separately on each channel to obtain the histogram of each color channel .
Parameters ：
image： Array , The input image .
nbins： Integers , Optional , Used to calculate the histogram bin Number . For integer arrays , This value will be ignored .
source_range： character string , Optional ,‘image’( Default ) Determine the range of the input image . ‘dtype’ Determine the expected range of the data type image .
normalize： Boolean type , Optional , If True, The histogram is normalized by the sum of its values .
Return value ：
hist： Array , Histogram value .
bin_centers： Array ,bin The value of the center .

2、 Returns the cumulative distribution function of a given image (cdf).

cucim.skimage.exposure.cumulative_distribution(image, nbins=256)

Parameters ：
image： Array , Image array .
nbins： Integers , Optional , Image histogram of bin Number .
return ：
img_cdf： Array , The value of the cumulative distribution function .
bin_centers： Array , The center of the box .

3、 Histogram matching (histogram matching)
meaning ： Make the cumulative histogram of the source image consistent with the target image
from skimage.exposure import match_histograms
Parameters 1： The source image ; Parameters 2： Target image ; Parameters 3： Multi channel matching

matched = match_histograms(image, reference, multichannel=True)