This is part 3.3 of the Geodata Management Course 

Here is a **detailed English summary** of the presentation *“Georeferencing in QGIS”* by Thomas Bauer & Franz Suppan, including references to the relevant slides:

---

### **Georeferencing in QGIS – Summary**

*(Slides 1–3)*

**Georeferencing** is the process of aligning geographic data (like raster images or scanned maps) to a known coordinate system. This enables the data to be viewed, queried, and analyzed alongside other spatial datasets. The main tasks include shifting, rotating, scaling, skewing, and sometimes more complex techniques like warping, rubber sheeting, or orthorectification.

Goals:

* Correct image distortions or degradation.
* Fit the image into a known map projection.

Georeferencing typically involves two main steps:

1. **Geometric Transformation**
2. **Resampling**

---

### **Methodology**

*(Slide 4)*

#### **Transformation**

This step assigns the correct geometric location to each pixel of the image.

* **Parametric Transformation**: Based on image acquisition geometry (not supported in QGIS).
* **Non-Parametric Transformation**: Uses ground control points (GCPs) and includes:

  * Linear transformation
  * Helmert transformation
  * Polynomial transformation
    (Note: No Digital Elevation Model (DEM) required)

Prerequisite: Ground control points with known coordinates in both the image and the reference system.

---

### **Types of Transformations**

*(Slides 5–6)*

Transformation types depend on the number of GCPs:

* **Linear** (min. 2 GCPs)
* **Helmert**
* **Polynomial** (min. 3 GCPs)
* **Thin Plate Spline**
* **Projective** (min. 8 GCPs)

Equations are used to compute transformation functions (example equations provided on Slide 5).

---

### **Affine Transformation**

*(Slide 6)*

A **first-order (affine) transformation** can:

* Shift
* Scale
* Rotate a raster

Requires at least **three links** (GCPs). More than three links may introduce residual errors. Higher-order transformations (second, third, etc.) allow more complex corrections but also require more points and processing time.

---

### **Resampling / Interpolation**

*(Slides 7–8)*

This step recalculates the pixel values (e.g., gray values) after transformation.

Approaches:

* **Nearest Neighbour**
* **Bilinear Interpolation**
* **Higher-order Interpolation** (e.g., Cubic Convolution)

The choice impacts output image quality and accuracy.

Slide 8 includes a visual comparison of **bilinear vs. nearest neighbour** interpolation.

---

### **Ground Control Points (GCPs)**

*(Slide 9)*

Good GCPs are critical for accurate transformation. Requirements:

* Precisely identifiable
* Stable (non-moving) objects
* High contrast
* Clear contours
* Appropriate size (related to image resolution)
* Evenly distributed across the image

Visual examples show good (😊) and poor (☹️) GCPs.

---

### **QGIS – Georeferencer Tool**

*(Slides 10–15)*

**Georeferencer GDAL plugin** is used in QGIS for georeferencing raster images.

Basic procedure:

1. Open the Georeferencer tool.
2. Click on a GCP in the image.
3. Click on the matching GCP in the reference dataset.
4. Repeat for multiple points.

Helpful tips:

* Use *View → Panels → GCP table* to monitor points and residuals.
* If residuals are not visible, reset the transformation type under *Transformation Settings*.
* Aim for a residual error of **1–1.5 pixels** for acceptable accuracy.

---

