Segmentation and Morphology

CMSC 178IP - Module 08

Noel Jeffrey Pinton
Department of Computer Science
University of the Philippines Cebu

Segmentation and Morphology

CMSC 178IP - Module 08

Noel Jeffrey Pinton
Department of Computer Science
University of the Philippines Cebu

Learning Objectives

By the end of this module, you will be able to:

Apply thresholding techniques (global, Otsu, adaptive)
Implement region-based segmentation methods
Use the watershed algorithm for complex segmentation
Apply morphological operations (erosion, dilation, opening, closing)
Use morphology for noise removal and object extraction

Thresholding

Converting grayscale to binary

Global Thresholding

Binary Thresholding:

g(x,y) = \begin{cases} 1 & \text{if } f(x,y) > T \\ 0 & \text{otherwise} \end{cases}

Simple but requires manual threshold selection.

Otsu's Thresholding

Otsu's Method: Automatically find optimal threshold by maximizing between-class variance.

\sigma^2_B = \omega_0 \omega_1 (\mu_0 - \mu_1)^2

Works well for bimodal histograms.

Knowledge Check

Think About It

When does Otsu's method work well, and when might it fail?

Click the blurred area to reveal the answer

Adaptive Thresholding

Adaptive (Local) Thresholding:
Compute threshold for each pixel based on local neighborhood. Handles uneven illumination.

Region-Based Segmentation

Grouping similar pixels

Region Growing

Region Growing Algorithm:

Start from seed point(s)
Examine neighboring pixels
Add if similar (within threshold)
Repeat until no more pixels can be added

Watershed Segmentation

Watershed: Treat gradient image as topography. "Flood" from markers - watershed lines form where waters meet. Excellent for separating touching objects.

Knowledge Check

Think About It

What is over-segmentation in watershed, and how can it be addressed?

Click the blurred area to reveal the answer

Morphological Operations

Shape-based image processing

Structuring Elements

Structuring Element: Small shape used as a probe. Common shapes: square, disk, cross, diamond.

Erosion and Dilation

Erosion

A \ominus B

Shrinks objects, removes small protrusions

Dilation

A \oplus B

Expands objects, fills small holes

Opening and Closing

Opening

A \circ B = (A \ominus B) \oplus B

Removes small objects, smooths contours

Closing

A \bullet B = (A \oplus B) \ominus B

Fills small holes, connects nearby objects

Knowledge Check

Think About It

What is the difference between erosion and opening?

Click the blurred area to reveal the answer

Morphological Gradient

Morphological Gradient:

G = (A \oplus B) - (A \ominus B)

Dilation minus erosion = edge detection using morphology

Grayscale Morphology

Grayscale Morphology: Same operations applied to grayscale images. Useful for top-hat transforms and contrast enhancement.

Applications

Practical uses of segmentation and morphology

Noise Removal Pipeline

Morphological Filtering: Opening removes salt noise (white specks), closing removes pepper noise (black specks).

Edge Detection with Morphology

Comparing morphological gradient with other edge detection methods

Object Extraction

Connected Component Analysis: Label distinct objects after segmentation. Extract properties: area, centroid, bounding box.

Texture Enhancement

Top-hat and bottom-hat transforms for texture analysis

Multi-Scale Segmentation

Different structuring element sizes capture features at different scales

Implementation

import cv2
import numpy as np

img = cv2.imread('image.jpg', 0)

# Otsu's thresholding
_, binary = cv2.threshold(img, 0, 255, 
    cv2.THRESH_BINARY + cv2.THRESH_OTSU)

# Adaptive thresholding
adaptive = cv2.adaptiveThreshold(img, 255,
    cv2.ADAPTIVE_THRESH_GAUSSIAN_C, cv2.THRESH_BINARY, 11, 2)

# Morphological operations
kernel = cv2.getStructuringElement(cv2.MORPH_ELLIPSE, (5,5))
opened = cv2.morphologyEx(binary, cv2.MORPH_OPEN, kernel)
closed = cv2.morphologyEx(binary, cv2.MORPH_CLOSE, kernel)

Summary

Key Takeaways

Thresholding: Global, Otsu (automatic), adaptive (local)
Region growing: Expand from seeds based on similarity
Watershed: Separate touching objects using markers
Erosion/Dilation: Basic morphological operations
Opening/Closing: Remove noise, fill holes
Applications: Noise removal, edge detection, object extraction

Questions?

Thank you for your attention!

Next: Module 09 - Computer Vision & Deep Learning I

End of Module 08

Segmentation and Morphology

Questions?