Utilizing WaveFunctionCollapse Algorithm for Procedural Generation of Terrains using Remotely Sensed Elevation Data

TL;DR

This paper adapts the WaveFunctionCollapse algorithm to generate realistic terrain height maps from remotely sensed elevation data by using slopes, effectively capturing structural features for virtual landscape creation.

Contribution

It introduces a novel approach of using slope-based input and height transformations with WFC for terrain generation from SRTM data, enhancing realism.

Findings

WFC effectively preserves input terrain structure.

Slope-based input improves terrain realism.

Generated terrains match statistical properties of real data.

Abstract

Procedural terrain generation plays a vital role in creating virtual landscapes for games, simulations, and various applications. The WaveFunctionCollapse (WFC) algorithm has proven effective in generating content by learning patterns from example data. In this research, we adapt WFC to generate terrain height maps using Shuttle Radar Topography Mission (SRTM) data. Instead of directly using raw height values, we use slopes to better capture structural features and preserve terrain patterns. Statistical comparisons, including histogram analysis, as well as evaluations of the mean, median, and standard deviation of input and output data, demonstrate that the algorithm effectively retains the input's structural characteristics while generating new terrain. the results show that WFC, with slope-based input and height-level transformations, can generate realistic terrain patterns for applications in game development and beyond.