Scientific-Visualization

This repository contains assignments for a course in Scientific Visualization, implemented in C and OpenGL. It demonstrates several visualization techniques including marching cubes, ray casting, Line Integral Convolution, Sammon mapping, and Self-Organizing Maps (SOM).

Table of Contents

• Overview
• Homework Assignments
• How to Run

Overview

This project explores several scientific visualization techniques through a series of assignments:

• HW1: Iso-surface — Marching Cubes — Extracting iso-surfaces from volumetric data.
• HW2: Ray Casting — Direct volume rendering using ray tracing techniques.
• HW3: Line Integral Convolution (LIC) — Visualizing vector fields with texture-based methods.
• HW4: Sammon Mapping — Mapping scalar fields onto surfaces.
• HW5: Self-Organizing Maps (SOM) — Dimensionality reduction and clustering visualization.

Homework Assignments

Hw1: Iso-surface Extraction (Marching Cubes)

Implemented the Marching Cubes algorithm to extract iso-surfaces from volumetric data.

View Report

Key Techniques

• Support for multiple iso-values and line/fill rendering modes
• Camera control and normal vector calculation options
• Cross-section and slice visualization features
• Histogram panel for iso-value selection with log2 transformation

Main Interface Overview

Histogram Panel

• Allows switching between displaying the Original Histogram or the Log2-transformed Histogram.
  The log2-transformed histogram makes it easier to identify appropriate iso-value regions.
• Each dataset has a predefined iso-value, when Line Mode and Multiple Iso-surface are enabled, the colors of the lines on the histogram correspond to the iso-values of the surfaces.

Control Panel

• Adjust the camera position.
• Choose between two normal vector computation methods: Triangle Normals or Gradient Normals.
• Select between two polygon modes: Fill Mode or Line Mode.
• Enable Multiple Iso-surface display, and choose which surfaces to show (up to 3 surfaces; you can display one, two, or all three simultaneously).
  Line Mode provides clearer visualization of the surfaces.
• Enable Cross Section, which offers two functions: Cross Section or Slice. You can select the direction of the cross section or slice and adjust its position.

File Panel

• Select different datasets to visualize.
  Switching datasets takes significant time due to recomputing the triangles.

Results

Iso-surface Animation

Carp Cross Section

Engine Cross Section (Left) / Slice (Right)

Aneurism

Hw2: Volume Rendering using Ray Casting (GPU Accelerated)

Implemented direct volume rendering using ray casting accelerated by GPU.

View Report

Key Techniques

• Adjustable ray step gap for performance/quality tradeoff
• Phong shading mode support
• Interactive camera manipulation
• Color transfer function editor for dynamic visualization

Main Interface Overview

Histogram Panel

File Panel

Control Panel

• Adjust the camera position.
• Adjust the Ray Casting Gap Size, which controls the sampling interval during ray tracing.
• Enable or disable Phong Shading Mode for enhanced surface shading.

Transfer Panel

• Configure the Transfer Function for color and opacity mapping.
• Add new color segments by inputting colors and defining ranges.
• Remove the most recently added color segment.

Results

Add Color Segments on Engine

Different Gap on Fish

Hw3: Line Integral Convolution (LIC) for Vector Field Visualization

Visualized 2D vector fields by convolving noise textures along streamlines using LIC.

View Report

Key Techniques

• Support for different noise types: black & white, gray-scale, spot noise
• Multiple convolution kernels: box filter, tent filter
• Color mode based on relative velocity magnitude
• Side-by-side display of noise and LIC results

Main Interface Overview

Control Panel

• Select different Noise Types.
• Select different Convolution Filter.
• Adjust the Convolution Mask Size(Convolution Kernel Size).
• Enable Color Display Mode, where colors represent relative velocity magnitudes.

File Panel

Display Panel

• The left panel shows the visualization result.
• The right panel shows the noise texture.

Results

vec13 — High-Frequency Noise: Black & White (Left) / Gray (Right), Different Noise Types

vec13 — Low-Frequency Noise: Box Filter (Left) / Tent Filter (Right), Different Convolution Filters

vec13 — Spot Noise

Other Data Results

Hw4: Sammon Mapping

Dimensionality reduction and visualization using Sammon mapping.

View Report

Results

HW5: Self-Organizing Maps (SOM)

Implemented Self-Organizing Map (SOM) trained on cylindrical lattices to generate a vase-shaped structure.

View Report

Key Techniques

• Adaptive learning rate
• Dynamic neighborhood layers decreasing over iterations

Training Parameters

The SOM training process uses the following parameters and rules:

1. Learning Rate

   The learning rate at each iteration is computed as:

   delta = 0.01 * exp(-iteration / max_iteration)

2. Neighborhood Layers

   The number of neighborhood layers t depends on the current iteration:

   if iteration > max_iteration / 2 then t = 1
   
   else if iteration > max_iteration / 5 then t = 2
   
   else if iteration > max_iteration / 10 then t = 3
   
   else t = 4
   

3. Neighborhood Function

   The influence of a node at distance d is computed as:

   r(d) = 1.0 / sqrt(d)

Results

How to Run

Development Environment

• Visual Studio 2022 or later
• OpenGL 3.3 or higher support

Folder Structure

• header/
  Contains project-specific header files such as shader.h, glad.h, and imgui.h.

• include/
  Contains third-party libraries headers including GLFW, KHR, glad, and glm.

• lib-vc2022/
  Contains precompiled .lib files required for linking (compatible with Visual Studio 2022).

Instructions

1. Open Visual Studio and select Open Folder, then choose this project’s root directory.
2. In Visual Studio, add header/ and include/ folders to the Additional Include Directories in the project properties.
3. Make sure OpenGL and GLFW libraries are installed on your system. Link the necessary .lib files in the Linker settings.
4. Build and run the project.
5. Each homework assignment is located in a separate folder; open and run the desired assignment accordingly.