Introduction to Computer Vision

• 1. What is Computer Vision?
  • 1.1 The “First Principles” Philosophy
• 2. The Human Visual System: Biology, Fallibility, and Ambiguity
  • 2.1 The Biological Pathway
  • 2.2 Qualitative vs. Quantitative Vision
  • 2.3 Case Studies in Visual Illusions
• 3. Topics Covered: Roadmap
  • Module Breakdown
• 4. Global Applications: Computer Vision in the Modern World

1. What is Computer Vision?

Computer vision is not merely a subset of artificial intelligence; it is a profound multidisciplinary engineering and scientific enterprise. It bridges the physical world and symbolic understanding, drawing upon optics, signal processing, electrical engineering, and computer science.

Vision pipeline: light source → scene → camera → Vision Software produces a scene description

The fundamental challenge lies in transforming raw numerical arrays—pixel data—into a meaningful description of the 3D environment.

Raw visual input: two children showering

Same scene as a numerical pixel matrix

In this field, the definition of the mission often dictates the methodology. Below is a comparison of the three primary philosophical pillars of computer vision research:

1.1 The “First Principles” Philosophy

While contemporary deep learning provides powerful tools, a First Principles approach—focusing on mathematical and physical underpinnings—is the prerequisite for generalizable and explainable AI. Relying on “black box” models often bypasses the structural understanding required for true innovation.

Why First Principles? Physical phenomena can be described with elegant math, making massive datasets and exhaustive training cycles unnecessary.

We prioritize these fundamentals for four reasons:

1. Precision and Conciseness — Physical phenomena can often be described with elegant math, rendering massive datasets unnecessary.
2. Debugging and Diagnostics — When a vision system fails, first principles provide the only rigorous framework for diagnosing the failure.
3. Synthetic Data Generation — When real-world data collection is impractical or dangerous, mathematical models allow us to synthesize high-fidelity training data.
4. Scientific Curiosity — The innate human drive to understand the “why” behind visual phenomena leads to breakthroughs that purely data-driven methods overlook.

2. The Human Visual System: Biology, Fallibility, and Ambiguity

Studying the human eye is a necessary starting point for designing artificial vision. Even though machines and humans often have divergent goals—qualitative navigation versus quantitative measurement—the eye provides a blueprint for efficient information reduction.

2.1 The Biological Pathway

The human visual system is a complex hierarchy designed for rapid analysis:

flowchart LR
    A["👁️ Eye & Lens<br/><i>Primary optical stage</i>"] --> B["🧬 Retina<br/><i>Early processing + data reduction</i>"]
    B --> C["🔌 Optic Nerve<br/><i>High-speed conduit</i>"]
    C --> D["🧠 LGN<br/><i>Relay station, directs to regions</i>"]
    D --> E["🎯 Visual Cortex<br/><i>Shape, color, motion, texture</i>"]
    
    style A fill:#1a1a2e,stroke:#e94560,color:#fff
    style B fill:#16213e,stroke:#e94560,color:#fff
    style C fill:#0f3460,stroke:#e94560,color:#fff
    style D fill:#1a1a2e,stroke:#e94560,color:#fff
    style E fill:#16213e,stroke:#e94560,color:#fff

Biological vision pathway: retina → LGN → visual cortex (V1, V2, MT/V5, V8)

2.2 Qualitative vs. Quantitative Vision

As engineers, we must recognize that human vision is a qualitative system, whereas factory automation, medical imaging, and robotics require quantitative precision. A human can recognize a face instantly but cannot measure the length of a component to the nearest millimeter. For tasks requiring extreme reliability, emulating human biology is often the “wrong” goal; machines must provide the measurable accuracy that biological systems lack.

2.3 Case Studies in Visual Illusions

Human vision is more fallible than it appears, often relying on internal assumptions to resolve ambiguity.

Example — Dongary Wave Illusion: The static leaf pattern below appears to shimmer or move due to involuntary micro-saccades of the eye.

Static leaf pattern that produces a perception of motion — the Dongary Wave illusion

Key Insight: While humans think through their visual experiences, machines must first learn to calculate through the rigorous lens of radiometry and geometry.

3. Topics Covered: Roadmap

This specialization covers the entire pipeline — from pixels to perception — organized into six modules:

flowchart TB
    subgraph Foundations["🟦 Foundations"]
        direction TB
        A["Introduction<br/><i>What is CV, human vision</i>"]
        B["Imaging<br/><i>Formation, sensors, processing</i>"]
    end
    
    subgraph Features["🟧 Features & 2D"]
        C["Features<br/><i>Edges, SIFT, stitching, faces</i>"]
    end
    
    subgraph Reconstruction["🟩 3D Reconstruction"]
        D["Reconstruction I<br/><i>Radiometry, photometric stereo</i>"]
        E["Reconstruction II<br/><i>Stereo, optical flow, SfM</i>"]
    end
    
    subgraph Perception["🟥 Perception"]
        F["Perception<br/><i>Tracking, segmentation, NN</i>"]
    end
    
    A --> B --> C --> D --> E --> F
    
    style A fill:#1a1a2e,stroke:#4cc9f0,color:#fff
    style B fill:#1a1a2e,stroke:#4cc9f0,color:#fff
    style C fill:#1a1a2e,stroke:#f72585,color:#fff
    style D fill:#1a1a2e,stroke:#06d6a0,color:#fff
    style E fill:#1a1a2e,stroke:#06d6a0,color:#fff
    style F fill:#1a1a2e,stroke:#e94560,color:#fff
    style Foundations fill:transparent,stroke:#4cc9f0,color:#4cc9f0
    style Features fill:transparent,stroke:#f72585,color:#f72585
    style Reconstruction fill:transparent,stroke:#06d6a0,color:#06d6a0
    style Perception fill:transparent,stroke:#e94560,color:#e94560

Module Breakdown

4. Global Applications: Computer Vision in the Modern World

Vision has evolved from a laboratory curiosity into a thriving global industry across diverse sectors.