Intro to digital images – Global BioImaging Data Course - Pune, India¶
Student version: 
Solution:
Welcome to the digital image chapter!
Just like the previous notebook, it is an interactive book: read, explore, and do.
# /// script
# requires-python = ">=3.11"
# ///
from pathlib import Path
from datetime import datetime
import numpy as np
from PIL import Image
import imageio
import skimage
import pooch
import matplotlib.pyplot as plt
%matplotlib inline
Student version:
Solution:
Welcome to the digital image chapter!
Just like the previous notebook, it is an interactive book: read, explore, and do.
Why Images?
Digital microscopy, image analysis, and machine learning all start with pixels.
Understanding that an image is just pixel values unlocks a world of quantitative analysis.
| Chapter | Concept | Why it matters |
|---|
| 0 | Importing modules | Using the Python ecosystem | 1 | Pixels as Arrays | Images are numbers you can compute on | | 2 | Loading & Display | Bring real data into NumPy, visualize it | | 3 | Image shape, dtype and channels | Dimensions, channels, and intensity histograms | | 4 | Image filtering and enhancement | Gaussian blur, edge detection | | 5 | File handlings and paths | Persist your results | | 6 | Mini project | Exploring a multi-channel fluorescence microscopy image |
Concept.
import pulls in modules – files containing variables, functions, classes.
Benefits:
Pro tip: use virtual environments (venv, conda) to isolate dependencies.
Example:
Import datetime and print today’s date. Hiint: use the today() function from date.
2025-06-16
Concept: Image representation as numerical arrays
A digital image is nothing more than a rectangular grid of pixels.
Each pixel stores one or more numbers:
| Image type | Numbers per pixel | Typical NumPy shape |
|---|---|---|
| Grayscale | 1 (pixel intensity: 0–255 8-bit, 0-4095 12-bit, 0-65535 16-bit) | (height, width), 2D array |
| RGB color | 3 (R, G, B) | (height, width, 3), 3D array |
Because the grid is regular, it maps perfectly to a NumPy array.
Libraries we will use:
numpy – numerical processingmatplotlib – visualizationskimage – image processingTiny 3 × 3 example
import numpy as np
img = np.zeros((3, 3), dtype=np.uint8)
print(img)
print("Shape:", img.shape)
[[0 0 0]
[0 0 0]
[0 0 0]]
Shape: (3, 3)
With NumPy you can create, inspect, and transform images with ordinary array syntax (add, multiply, slice…). These tiny examples scale up seamlessly to 1024 × 1024 microscopy images.
Then, import pyplot and image modules from the ploting library matplotlib. Using it, we can display our checkerboard array in image form:
Create a 5 × 5 NumPy array called blink where the center pixel has value 255 and the rest are 0.
Print the array and its shape.
Concept.
Real images live on disk as PNG/JPEG/TIFF files, and many more formats! Bioformats is a library that can load many of them. Here, we'll load an image from a URL with imageio, convert to a NumPy array with np.asarray, and plot with matplotlib.
Let's crop the image and inspect the pixel values:
Basic operations on images
Viewing Properties
Cropping and slicing
Image inversion
Rescaling intensity
Locate a cell by slicing the image, then print the values and the patch.
Hint: there is a cell around (100, 200)
Adapt the code above to plot the composite of channels 0 and 1.
Hint: you can use np.stack to create the composite image.
Concept: once an image is a NumPy array, vectorized math lets you transform it lightning‑fast:
Why it matters: Shape and dtype affect how we process and visualize the image.
Plot the pixel intensity histogram of the image:
A histogram shows how often each pixel value occurs.
Useful to:
Smoothing with Gaussian blur
Gaussian blur is a fundamental image processing technique that:
The Gaussian function gives more weight to nearby pixels and less weight to distant ones, creating a natural-looking blur effect.
x = np.linspace(-3, 3, 20)
y = np.linspace(-3, 3, 20)
X, Y = np.meshgrid(x, y)
Z = np.exp(-(X**2 + Y**2) / 2) / (2 * np.pi)
plt.imshow(Z, cmap='gray')
plt.show()
Edge detection with Sobel operator
The Sobel operator is a popular edge detection algorithm that:
Concept. Files let your programs persist data.
open(path, mode) inside a context manager (with) to ensure automatic close.'r' read, 'w' write (truncate), 'a' append, 'b' binary, '+' read/write.Best practice: work with paths using pathlib.Path.
from pathlib import Path
import time
path = Path('demo.txt') # will create the file if it doesn't exist
with path.open('w') as f:
f.write(f'first line {datetime.now()}\n')
time.sleep(1)
f.write(f'second line {datetime.now()}')
with path.open('r') as f:
data = f.read()
# print line by line
for line in data.split('\n'):
print(line)
first line 2025-06-14 13:50:56.601441 second line 2025-06-14 13:50:57.605721
Load image from a file:
Accessing metadata:
Smooth the image and save it to disk.
Hint: use skimage.io.imsave to save the image.
To check the result, you can load the image and display it:
In this mini project, we will:
First, let's load the image and examine its structure:
Save the image to disk for later use:
Introduction to napari
napari is a fast, interactive, multi-dimensional image viewer for Python. It supports:
How to install napari:
For now, we will install napari using the asset system as a bundled app.
Go to https://github.com/napari/napari/releases/tag/v0.6.1, and install the appropriate version for your operating system.