Bitmap Images in Computer Vision

🖼️ What Is a Bitmap Image?

A bitmap image represents an image as a grid of pixels,
where each pixel stores explicit intensity or color values.

Key properties:

• Discrete spatial sampling
• Discrete intensity values
• Direct memory mapping

In computer vision, bitmap images are treated as numerical matrices.

---

📐 Bitmap as a Matrix

A bitmap image of size \(H \times W\) can be written as:

• Single-channel:
  \(I \in \mathbb{R}^{H \times W}\)

• Multi-channel:
  \(I \in \mathbb{R}^{H \times W \times C}\)

Where:

• height \(H\)
• width \(W\)
• number of channels \(C\)

---

⚫ Grayscale Image (1 Channel)

Representation

Each pixel stores one intensity value.

\[I(x,y) \in [0, 255]\]

(8-bit grayscale)

Example

0   → black
255 → white

Characteristics

• Simple
• Memory efficient
• Most classical CV algorithms operate on grayscale

Typical Usage

• Edge detection
• Thresholding
• Geometry (Hough, LSM)

---

🎨 Multi-Channel Images

3-Channel (RGB / BGR)

Each pixel has three values:

\[I(x,y) = [R, G, B]\]

Each channel: \(R, G, B \in [0, 255]\)

Characteristics

• Encodes color information
• Higher memory cost
• Often converted to grayscale for processing

Typical Usage

• Visualization
• Color-based segmentation
• Feature extraction

---

🧠 N-Channel Images (General Case)

In computer vision, channels are not limited to color.

\[I(x,y) = [c_1, c_2, \dots, c_N]\]

Examples:

• RGB → 3 channels
• RGBA → 4 channels
• Hyperspectral → dozens or hundreds of channels
• Feature maps in CNNs

---

🌈 Common Channel Types

Channels	Meaning	Usage
Gray	Intensity	Geometry, filtering
RGB	Color	Visualization
HSV	Hue/Saturation	Color segmentation
Depth	Distance	3D vision
Normal	Surface orientation	3D reconstruction
Feature	Learned	Deep learning

---

📦 Memory Layout (Important)

Interleaved (HWC)

[R G B][R G B][R G B]

Planar (CHW)

RRR...
GGG...
BBB...

Layout matters for:

• Performance
• SIMD / GPU access
• Library compatibility

---

🔢 Bit Depth

Pixel values can have different precision:

Bit Depth	Range
1-bit	0–1
8-bit	0–255
16-bit	0–65535
Float	Real values

Higher bit depth:

• More dynamic range
• Higher memory cost

---

⚖️ Grayscale vs Multi-Channel

Aspect	Grayscale	Multi-Channel
Memory	Low	High
Speed	Fast	Slower
Information	Intensity only	Rich
CV usage	Core algorithms	Specialized tasks

---

🧠 Practical Insight

In classical computer vision:

Input (RGB)
   ↓
Convert to Grayscale
   ↓
Statistics / Geometry
   ↓
Decision

Color is often auxiliary, not primary.

---

🎯 Takeaway

Bitmap images are structured numeric data, not pictures.

Understanding:

• Channels
• Bit depth
• Memory layout

is essential for:

• Performance
• Correct processing
• Algorithm design 🚀