How to view the pixel values of an image using scikit-learn in Python?

Viewing pixel values of an image is a fundamental step in image processing and computer vision tasks. Scikit-image provides convenient functions to read images and extract pixel data, which can then be converted to a pandas DataFrame for analysis.

Images are stored as multi-dimensional arrays where each pixel has intensity values. For RGB images, each pixel contains three values (Red, Green, Blue), while grayscale images have single intensity values per pixel.

Reading and Displaying an Image

First, let's read an image and display its basic properties ?

from skimage import io, data
import pandas as pd
import numpy as np

# Using a sample image from scikit-image
img = data.camera()  # Grayscale image
print("Image shape:", img.shape)
print("Image data type:", img.dtype)
print("Min pixel value:", img.min())
print("Max pixel value:", img.max())

Image shape: (512, 512)
Image data type: uint8
Min pixel value: 0
Max pixel value: 255

Converting Image Pixels to DataFrame

The flatten() method converts the 2D image array into a 1D array, making it easy to create a DataFrame ?

from skimage import data
import pandas as pd

# Load sample image
img = data.camera()
print("Original image dimensions:", img.shape)

# Flatten the image to get pixel values in 1D array
pixel_values = img.flatten()
print("Flattened array length:", len(pixel_values))

# Create DataFrame from pixel values
df = pd.DataFrame(pixel_values, columns=['Pixel_Value'])
print("\nDataFrame shape:", df.shape)
print("\nFirst 10 pixel values:")
print(df.head(10))

Original image dimensions: (512, 512)
Flattened array length: 262144

DataFrame shape: (262144, 1)

First 10 pixel values:
   Pixel_Value
0          162
1          162
2          162
3          162
4          162
5          163
6          163
7          163
8          163
9          164

Working with Color Images

For RGB images, each pixel has three color channels. We can create separate columns for each channel ?

from skimage import data
import pandas as pd

# Load a color image
img_color = data.astronaut()
print("Color image shape:", img_color.shape)

# Reshape to get each pixel as a row with RGB values
pixels = img_color.reshape(-1, 3)
print("Reshaped pixels array shape:", pixels.shape)

# Create DataFrame with separate columns for R, G, B
df_color = pd.DataFrame(pixels, columns=['Red', 'Green', 'Blue'])
print("\nColor DataFrame shape:", df_color.shape)
print("\nFirst 5 rows:")
print(df_color.head())

# Basic statistics
print("\nPixel value statistics:")
print(df_color.describe())

Color image shape: (512, 512, 3)
Reshaped pixels array shape: (262144, 3)

Color DataFrame shape: (262144, 3)

First 5 rows:
   Red  Green  Blue
0   61     55    123
1   62     56    124
2   63     57    125
3   65     59    127
4   67     61    129

Pixel value statistics:
               Red        Green         Blue
count   262144.0000   262144.0000   262144.0000
mean       119.2767      89.8135      139.0453
std         58.9895      49.1469       67.8479
min          0.0000       0.0000        0.0000
25%         73.0000      51.0000       89.0000
50%        118.0000      83.0000      137.0000
75%        163.0000     124.0000      186.0000
max        255.0000     255.0000      255.0000

Adding Pixel Position Information

Sometimes it's useful to include the pixel coordinates along with their values ?

from skimage import data
import pandas as pd
import numpy as np

# Load grayscale image
img = data.coins()[0:100, 0:100]  # Take smaller portion for demonstration
height, width = img.shape

# Create coordinate arrays
y_coords, x_coords = np.meshgrid(range(height), range(width), indexing='ij')

# Create DataFrame with coordinates and pixel values
df_with_coords = pd.DataFrame({
    'x': x_coords.flatten(),
    'y': y_coords.flatten(), 
    'pixel_value': img.flatten()
})

print("DataFrame with coordinates:")
print(df_with_coords.head(10))
print(f"\nTotal pixels: {len(df_with_coords)}")

DataFrame with coordinates:
   x  y  pixel_value
0  0  0          170
1  1  0          170
2  2  0          170
3  3  0          170
4  4  0          170
5  5  0          170
6  6  0          170
7  7  0          170
8  8  0          170
9  9  0          170

Total pixels: 10000

Practical Applications

Converting image pixels to DataFrames enables various data science operations ?

from skimage import data
import pandas as pd

# Load image and create DataFrame
img = data.camera()
df = pd.DataFrame(img.flatten(), columns=['intensity'])

# Find bright pixels (intensity > 200)
bright_pixels = df[df['intensity'] > 200]
print(f"Number of bright pixels: {len(bright_pixels)}")

# Find dark pixels (intensity < 50)
dark_pixels = df[df['intensity'] < 50]
print(f"Number of dark pixels: {len(dark_pixels)}")

# Calculate histogram-like distribution
intensity_counts = df['intensity'].value_counts().sort_index()
print(f"\nMost common intensity value: {intensity_counts.idxmax()}")
print(f"Occurs {intensity_counts.max()} times")

Number of bright pixels: 5982
Number of dark pixels: 21073

Most common intensity value: 0
Occurs 5476 times

Conclusion

Converting image pixels to pandas DataFrames using flatten() or reshape() enables powerful data analysis on image data. This approach is particularly useful for image preprocessing, statistical analysis, and preparing data for machine learning models.