Exercise: Thresholding

In this exercise, we will try out different threshold methods and see how they perform on the same data and create a label image from the binarized image data.

Recap

Segmentation is the process of separating single pixels into different categories that may or may not have specific meaning attached to that. Thresholding is a simple, yet often effecte segmentation technique. In thresholding, pixels are separated into two categories - the background and the foreground of an image - by comparing the intensities of each image to a single value, the threshold value. Finding a “good” threshold value is therefore of key importance.

The output of a thresholding operation is a binary image, consiting of 0s and 1s - which is equivalent to True and False in terms of boolean values.

Vocabulary: Thresholding is a type of segmentation. Separating the image into two types of pixels (background and foreground) is a type of semantic segmentation, and is - since only two types of pixels exist in the output - referred to as a binarization operation.

How to code: Remember - in Python you can compare objects (single numbers or images) to threshold values simply by using the < or > operator - make sure to put the output into a new variable!

from skimage import data, filters, measure
import matplotlib.pyplot as plt
import napari

image = data.human_mitosis()
plt.imshow(image, cmap='gray')

<matplotlib.image.AxesImage at 0x232307465e0>

Now, apply several of the thresholds from skimage to the image data! Choose one that you think is suitable. Hint: You can find them under filters.threshold_... - use the tab key or the documentation to find out how to use the implemented threshold functions.

binary_image = # code goes here

Again, use plt.imshow() to visualize the results

# Code goes here

Now we would like to perform connected-component analysis. Use the appropriate function from skimage (measure.label()) for this task!

label_image = # code goes here

Visualization in Napari

viewer = napari.Viewer()

WARNING: DirectWrite: CreateFontFaceFromHDC() failed (Indicates an error in an input file such as a font file.) for QFontDef(Family="8514oem", pointsize=12, pixelsize=20, styleHint=5, weight=50, stretch=100, hintingPreference=0) LOGFONT("8514oem", lfWidth=0, lfHeight=-20) dpi=240

Use the add_image() and the add_labels() function to add image, binary_image and label_image to the viewer and create screenshot of the viewer with the napari.utils.nbscreenshot() function.

# Code goes here