Digital Image Representation & Processing: Pixels, Formats, and Techniques

TAKEAWAYS:

• Digital images are represented as pixel grids (2D arrays of RGB values) with resolution measured in DPI/PPI.
• Image formats (JPEG, PNG, BMP) use lossy/lossless compression to balance quality and file size.
• Image processing techniques (filtering, morphing, segmentation) enable editing tools in apps like eSewa’s transaction receipts.
• Color models (RGB, CMYK, HSV) determine how images are displayed or printed.
• Compression (lossy vs. lossless) is critical for bandwidth (e.g., NTC’s live video feeds).
• Real-world applications range from medical imaging (X-rays) to e-commerce (Daraz product images).

1. Digital Image Representation: From Analog to Pixels

Digital images are mathematical representations of real-world scenes captured by sensors (e.g., camera CCDs). Unlike analog photos, they are stored as discrete pixel values in a grid.

How Images Are Digitized

1. Sampling: The analog scene is divided into a grid of pixels (picture elements).
   • Resolution is measured in DPI (dots per inch) or PPI (pixels per inch).
   • Higher DPI = sharper image but larger file size.
2. Quantization: Each pixel’s color is mapped to a finite number of values (e.g., 8 bits per channel = 256 shades).
3. Encoding: Pixel data is stored in a structured format (e.g., BMP, JPEG).

graph TD
    A["Analog Scene"] --> B["Camera Sensor"]
    B --> C["Sampling: Divide into Pixels"]
    C --> D["Quantization: Assign RGB Values"]
    D --> E["Encoding: Store as Binary Data"]

Pixel Grid and Color Depth

• A pixel is typically represented by 3 bytes (RGB) or 4 bytes (RGBA, including alpha transparency).
• Color depth determines the number of colors:
  • 8-bit: 256 colors (e.g., black & white).
  • 24-bit: 16.7 million colors (full color).
  • 32-bit: 16.7M colors + transparency.

Worked Example: Calculating File Size Problem: A 2000×3000 pixel image with 24-bit color. Solution:

• Total pixels = 2000 × 3000 = 6,000,000 pixels.
• Bytes per pixel = 3 (RGB).
• Total size = 6,000,000 × 3 = 18,000,000 bytes (≈17.15 MB).

2. Image File Formats: How Data Is Stored

Different formats optimize for compression, quality, or editing:

Why JPEG is Everywhere

• Uses lossy compression (discards "unimportant" data).
• Works well for photographic images (human eyes can’t detect small losses).
• Example: Daraz compresses product images to JPEG to reduce server load.

3. Image Processing Techniques

Processing alters pixel data for enhancement, editing, or analysis.

A. Basic Operations

1. Brightness/Contrast Adjustment
   • Changes pixel intensity (e.g., darkening a photo).
2. Color Space Conversion
   • RGB → HSV (for hue-based editing).
3. Geometric Transformations
   • Rotation, scaling, flipping (e.g., Pathao’s driver dashboard images).

B. Advanced Techniques

1. Filtering
   • Smoothing: Reduces noise (e.g., blurring a blurry camera shot).
   • Edge Detection: Highlights boundaries (e.g., medical X-rays).
2. Morphing
   • Smoothly transitions between two images (e.g., animated logos).
3. Segmentation
   • Separates objects from background (e.g., OCR in bank checks).

Worked Example: Edge Detection in Traffic Surveillance

Scenario: NTC uses Canny edge detection to identify lane markings in live traffic camera feeds. Steps:

1. Apply Gaussian blur to reduce noise.
2. Compute gradients (Sobel operator).
3. Threshold to highlight strong edges. Result: Clear lane detection for autonomous systems.

4. Color Models: How Computers Represent Color

Different models serve different purposes:

A 3D cube showing RGB primary colors and their mixtures. (Image: SharkD, CC BY-SA 4.0, via Wikimedia Commons)

RGB vs. CMYK: Why Printers Use CMYK

• RGB adds light (additive model).
• CMYK subtracts light (subtractive model).
• Example: A red shirt printed with CMYK uses 0% C, 100% M, 100% Y, 0% K.

5. Image Compression: Reducing File Size

Compression balances quality vs. size for storage/transmission.

A. Lossless Compression

• No data loss (e.g., PNG, ZIP).
• Methods:
  • Run-Length Encoding (RLE): Compresses repeated pixels (e.g., black & white scans).
  • Huffman Coding: Assigns shorter codes to frequent pixels.

B. Lossy Compression

• Discards "unimportant" data (e.g., JPEG).
• Methods:
  • DCT (Discrete Cosine Transform): Blocks of pixels are transformed into frequency components.
  • Quantization: Reduces precision of coefficients.

Worked Example: Compressing a Selfie for WhatsApp

Problem: A 4000×6000 pixel selfie (24-bit) is 72 MB. Solution:

1. Resize to 1000×1500 (reduces pixels by 4×).
2. Apply JPEG compression (quality = 80%).
3. Final size: ~5 MB (fits WhatsApp’s 10 MB limit).

6. Real-World Applications

In the Real World

1. Daraz Product Images

   • Idea: JPEG compression for fast loading.
   • How: Product photos are resized and compressed to <500 KB for mobile users.

2. Pathao Driver Dashboard

   • Idea: Real-time image processing (edge detection for lane marking).
   • How: Cameras feed live video to detect traffic rules violations.

3. NEPSE Stock Charts

   • Idea: SVG vector graphics for scalability.
   • How: Charts render crisply at any resolution (unlike raster images).

4. eSewa Transaction Receipts

   • Idea: PNG format for transparency (logo overlay).
   • How: Receipts use 24-bit PNG to display QR codes clearly.

Exam Tip

• Focus on:
  • Pixel representation (DPI, color depth, file size calculations).
  • Format differences (JPEG vs. PNG trade-offs).
  • Processing techniques (filtering, segmentation, morphing).
  • Color models (RGB for screens, CMYK for print).
  • Compression (lossy vs. lossless, DCT in JPEG).
• Common Pitfalls:
  • Confusing DPI (physical resolution) with PPI (digital resolution).
  • Forgetting that JPEG is lossy while PNG is lossless.
  • Not linking real-world apps (e.g., Daraz, Pathao) to techniques.
• Question Patterns:
  • Define terms (e.g., "What is DCT?").
  • Compare formats (e.g., "Why use PNG over JPEG for icons?").
  • Apply processing (e.g., "How would you detect edges in a traffic camera image?").

Final Note: Always visualize pixel grids, compression blocks, and color models in exams—draw them if needed!