← ArchivePaper2015
Real-Time Dynamic Wrinkling of Coarse Animated Cloth
Russell Gillette, Craig Peters, Nicholas Vining, Essex Edwards, Alla Sheffer
Real-time method adding plausible wrinkle details to coarsely simulated or keyframed cloth by using a data-driven wrinkle prediction model.
Abstract
We present a real-time method for adding dynamic, believable wrinkles to coarse, low triangle count cloth animation typical of video games, where no reference shape exists and the input animation is noisy. The approach uses a two stage stretch tensor estimation process: a graph-cut segmentation first detects spatially and temporally coherent compressing, stable, and stretching surface patches, and these motion patterns are then used to build a per-triangle temporally adaptive reference shape and stretch tensor. Wrinkle paths are traced across areas of high compression on the piecewise constant tensor field and evolved with a global optimization for temporal persistence. Wrinkle geometry and normals are generated on the fly using the GPU tessellation unit and fragment shader, sustaining 60 frames per second on current console and PC hardware.
How to read this
- Category
- Method: real-time wrinkle synthesis for coarse cloth
- Contributions
- Adds dynamic, believable wrinkles to coarse, low-triangle game cloth animation with no reference shape and noisy input.
- Uses a two-stage stretch-tensor estimation: graph-cut segmentation of coherent compressing/stable/stretching patches, then a per-triangle temporally adaptive reference shape and stretch tensor.
- Traces wrinkle paths across high-compression regions, enforces temporal persistence via global optimization, and generates geometry/normals on the GPU at 60 fps on console and PC hardware.
- Context
- A real-time, reference-free counterpart to example-based wrinkle methods such as Wang et al.'s Example-Based Wrinkle Synthesis for Clothing Animation.Builds on: Example-Based Wrinkle Synthesis for Clothing Animation
- Correctness
- Targets the hard case of noisy keyframed/coarse cloth with no rest reference, so wrinkles are plausible and temporally coherent rather than physically derived; quality depends on the stretch-tensor segmentation tracking the true compression regions.
- Clarity
- Moderately dense; a first pass conveys the pipeline stages, a second pass for the tensor estimation and wrinkle-path optimization.
- How to read it
- First pass for the two-stage tensor-to-wrinkle pipeline; second pass on the graph-cut segmentation and temporal optimization if implementing a real-time cloth-detail shader.
Builds on
Built upon by
Nothing yet.
Related work
- Adaptive Anisotropic Remeshing for Cloth Simulation 2012 / SIGGRAPH Asia
- Gravity Preloading for Maintaining Hair Shape Using the Simulator as a Closed-Box Function 2022 / SIGGRAPH
- Simulating Cloth Using Bilinear Elements 2021 / SIGGRAPH
- Dynamic Deformables: Implementation and Production Practicalities 2020 / SIGGRAPH Courses
Keywords
This page summarises the entry and links to its original source. The archive never hosts or redistributes the publication itself.Show it in the full archive list →