Module 1: Introduction to Computer Vision

1. What is Computer Vision?

Definition: Computer Vision is the field of study focused on enabling machines to process and interpret visual information from the world, similar to how humans do using their vision system.

• It enables automated tasks such as detection, classification, tracking, and recognition.
• Often powered by deep learning and image processing techniques.
• Used in diverse domains: healthcare, security, retail, robotics, and self-driving cars.

2. Key Concepts and Components

• Pixels: Smallest unit of an image; each pixel contains intensity/color information.
• Color spaces: RGB, HSV, Grayscale - various ways to represent color information.
• Image formats: JPG, PNG, TIFF - impact quality, compression, and storage.
• Resolution: The number of pixels in width and height of the image.

3. Applications of Computer Vision

• Autonomous Driving (lane and object detection)
• Medical Imaging (e.g., tumor detection, retinal scan analysis)
• Surveillance (e.g., person tracking, facial recognition)
• Retail (e.g., customer counting, shelf monitoring)
• OCR and historical document digitization

4. Image Understanding vs Image Processing

Image Processing: Basic transformations to improve or analyze an image (e.g., filters, enhancement).

Image Understanding: Extracting semantic meaning from visual data (e.g., classifying an image as a cat).

Image Processing → Feature Extraction → Image Understanding

5. Basic Equation of Image Formation

g(x, y) = f(x, y) * h(x, y) + n(x, y)

Where:
    - f(x, y): original image (ideal signal)
    - h(x, y): point spread function (camera optics)
    - n(x, y): additive noise
    - g(x, y): observed image captured by camera
    

This equation is a basic model for how digital images are formed considering system imperfections.

6. Visual Perception Pipeline

7. Real-World Example: Autonomous Vehicles

Self-driving cars use computer vision to interpret their surroundings.

• Input: Real-time camera feed from road
• CV Tasks: Detect lanes, traffic signs, pedestrians
• Output: Decisions for navigation, braking, or acceleration

8. Timeline of Key Milestones

• 1960s–80s: Rule-based and symbolic vision (edge detection, Hough transform)
• 1990s: Hand-crafted features (SIFT, SURF, HOG)
• 2012: Deep learning revolution (AlexNet on ImageNet)
• Now: Transformers, self-supervised learning, multimodal vision

9. Hands-on Practice (Colab Compatible)

OpenCV Canny Edge Detection on a user-uploaded image:

!pip install opencv-python-headless matplotlib

import cv2
import numpy as np
import matplotlib.pyplot as plt
from google.colab.patches import cv2_imshow
from google.colab import files

uploaded = files.upload()
img_path = next(iter(uploaded))

# Load and show original
img = cv2.imread(img_path)
cv2_imshow(img)

# Convert to grayscale
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
cv2_imshow(gray)

# Gaussian blur
blur = cv2.GaussianBlur(gray, (5, 5), 0)

# Canny edge detection
edges = cv2.Canny(blur, 100, 200)
cv2_imshow(edges)

cv2.imwrite("canny_edges.jpg", edges)
    

10. Assignment