IMAGE PROCESSING

BCA Semester 6 - Internals Exam Study Guide

Complete Coverage of All Important Topics

Table of Contents
• 1. Digital Image Processing
• 2. Image Filtering
• 3. Image Restoration
• 4. Image Segmentation
• 5. Morphological Operations

1. Digital Image Processing

Definition & Overview
• Digital image processing is the manipulation of digital images using computer
algorithms
• Converts images into digital signals that can be processed, enhanced, and analyzed
• Fundamental technique in computer vision, medical imaging, remote sensing, and
multimedia

Key Concepts
Pixel (Picture Element)
• Smallest unit of a digital image
• Contains intensity value (0-255 for grayscale; RGB values for color)
Image Resolution
• Spatial resolution: Number of pixels (e.g., 1920×1080)
• Intensity resolution: Number of bits per pixel (8-bit = 256 levels)
Color Models
• RGB: Red, Green, Blue - additive color model
• CMYK: Cyan, Magenta, Yellow, Black - subtractive model
• Grayscale: Single intensity channel
• HSV: Hue, Saturation, Value

Image Processing Pipeline

Stage Description
Acquisition Capturing image using cameras/sensors
Preprocessing Enhancing image quality, noise reduction
Segmentation Dividing image into meaningful regions
Feature Extraction Identifying important characteristics
Classification/Recognition Interpreting image content
Output Generating results or enhanced image
2. Image Filtering
Definition & Purpose
• Process of applying a filter kernel to an image to enhance or extract features
• Purpose: Noise reduction, edge detection, blurring, sharpening, enhancement

Linear Filters
Convolution
• Mathematical operation combining two functions to produce a third
• In images: Slide kernel over image, multiply overlapping values, sum results
• Formula: G(x,y) = ∑∑ F(i,j) × H(x-i, y-j)
Common Linear Filters
Filter Kernel Effect
Average/Mean All values = 1/9 Blur, noise reduction
Gaussian Gaussian curve Smooth blur, preserves edges
Sobel Gradient operators Edge detection
(horizontal/vertical)
Laplacian [-1,2,-1] Edge detection, enhancement
Box Filter All ones normalized Simple averaging blur

Non-Linear Filters
Median Filter
• Replaces each pixel with median of values in neighborhood
• Excellent for salt-and-pepper noise
• Preserves edges better than Gaussian blur
Bilateral Filter
• Edge-preserving smoothing filter
• Considers both spatial distance and intensity difference

Frequency Domain Filtering

• Use Fourier Transform to convert image to frequency domain
• Low-pass filter: Allows low frequencies (blurring)
• High-pass filter: Allows high frequencies (edge detection)
• Band-pass filter: Allows specific frequency range
3. Image Restoration
Definition
• Process of recovering a true image from a degraded image
• Degradation can be due to: noise, blur, geometric distortion
• Goal: Reconstruct original image with maximum quality

Types of Noise
Gaussian Noise
• Follows normal probability distribution
• Common in thermal and electronic cameras
Salt-and-Pepper Noise
• Random white and black pixels scattered in image
• Caused by: transmission errors, faulty pixels
Speckle Noise
• Multiplicative noise, common in radar/ultrasound
• Appears as grainy texture

Restoration Techniques
Mean Filter
• Simple averaging of neighboring pixels
• Good for Gaussian noise, blurs edges
Wiener Filter
• Optimal filter minimizing mean square error
• Requires knowledge of noise and signal statistics
• Better edge preservation than mean filter
Median Filter
• Non-linear filter, excellent for salt-and-pepper noise
• Preserves edges better than linear filters
Morphological Operations
• Erosion and Dilation for noise removal and edge detection
• Opening and Closing for smoothing

Deblurring
• Inverse Filtering: Divide frequency spectrum by blur kernel
• Constrained Least Squares: Iterative approach with smoothness constraint
• Blind Deconvolution: Unknown degradation filter

4. Image Segmentation
Definition & Importance
• Division of image into non-overlapping regions with similar properties
• Goal: Simplify image representation for analysis
 • Foundation for object detection, recognition, and scene understanding

Importance of Segmentation
Aspect Importance
Feature Extraction Isolate objects of interest for analysis
Object Recognition Identify and classify objects in scene
Medical Imaging Detect tumors, organs, abnormalities
Autonomous Vehicles Identify road, vehicles, pedestrians
Scene Understanding Separate foreground from background
Data Compression Reduce data by grouping similar regions

Segmentation Techniques
Thresholding
• Binary segmentation: Pixels above/below threshold
• Global thresholding: Single threshold for entire image
• Local thresholding: Different threshold for different regions
• Otsu's method: Automatic threshold selection
Edge-based Segmentation
• Detect edges using Sobel, Canny, Laplacian
• Canny Edge Detector: Optimal edge detection
• Hough Transform: Detect lines and circles
Region-based Segmentation
• Region Growing: Start from seed points, grow similar regions
• Split and Merge: Recursively divide/merge regions
• Watershed: Treats image as topographical map
Clustering-based Segmentation
• K-means Clustering: Partition pixels into K clusters
• Fuzzy C-means: Soft clustering, pixels can belong to multiple clusters
• Mean-shift: Mode-seeking algorithm
5. Morphological Operations
Definition
• Mathematical operations on binary/grayscale images based on shapes
• Tool from mathematical morphology for image analysis and processing
• Operates on image structure: shape, size, connectivity

Structuring Element
• Small shape/pattern used to probe and interact with image
• Common shapes: Square (3×3, 5×5), Circle, Cross, Diamond
• Choice affects operation outcome

Basic Operations
Erosion
• Operation: Output pixel = minimum of pixels under structuring element
• Effect: Shrinks white objects, removes small objects
• Use: Remove noise, separate objects
• Example: B ⊖ S (Image eroded by structuring element)
Dilation
• Operation: Output pixel = maximum of pixels under structuring element
• Effect: Expands white objects, fills holes
• Use: Fill gaps, connect objects
• Example: B ⊕ S (Image dilated by structuring element)

Compound Operations
Opening (Erosion → Dilation)
• Operation: (B ⊖ S) ⊕ S
• Effect: Remove small objects without changing overall shape
• Use: Noise removal, separating touching objects
Closing (Dilation → Erosion)
• Operation: (B ⊕ S) ⊖ S
• Effect: Fill holes inside objects
• Use: Connecting nearby objects, hole filling

Advanced Operations
Gradient (Dilation - Erosion)
• Extracts object boundaries
• Useful for edge detection
Top Hat
• Operation: B - Opening(B)
• Extracts small objects
Bottom Hat
• Operation: Closing(B) - B
• Extracts small holes
Applications of Morphological Operations
Application Operation Used Purpose
Noise Removal Opening Remove salt-and-pepper noise
Object Detection Gradient Find object boundaries
Image Enhancement Closing Fill holes and gaps
Object Separation Erosion Separate touching objects
Feature Extraction Skeleton Extract object structure
Image Analysis Connected Components Label and count objects

Key Points for Exam

Important Definitions
• Understand fundamental concepts: pixels, resolution, color models
• Know filter types: linear, non-linear, frequency domain
• Memorize noise types and restoration methods
• Understand segmentation importance and techniques
• Know morphological operations and their effects

Common Exam Questions

• What is digital image processing? Why is it important?
• Explain convolution. How is it used in filtering?
• Compare different types of noise and restoration techniques
• Explain image segmentation and its applications
• Define erosion and dilation. How do they work?
• What is opening and closing? Give examples
• Differentiate between edge-based and region-based segmentation

Study Tips
• Draw diagrams showing kernel operations and their effects
• Practice manual calculations: convolution, morphological operations
• Understand real-world applications for each topic
• Compare and contrast different techniques
• Review formulas: convolution, filter kernels, erosion/dilation
• Create flowcharts for image processing pipelines

Good Luck with Your Exam!