By-Example Synthesis of Vector Textures

TL;DR

The paper tackles the problem of generating scalable vector textures from a single raster exemplar by converting the raster to a three-layer vector representation comprising primary textons, secondary textons, and a background gradient field. It introduces an offline analysis phase (texton extraction, descriptor construction, and background summarization) followed by online synthesis (primary/secondary texton placement and background gradient generation), enabling arbitrarily large, editable vector textures. The main contributions include (1) a novel texton hierarchy and descriptor framework for raster-to-vector texture synthesis, (2) a practical synthesis pipeline with Poisson-disk placement and gradient-field interpolation, and (3) a set of enhancements and editing operations that leverage vector representations. The approach achieves competitive perceptual quality relative to raster baselines and offers editing capabilities that are difficult to realize with raster textures, highlighting the practical impact of direct vector texture synthesis from raster inputs.

Abstract

We propose a new method for synthesizing an arbitrarily sized novel vector texture given a single raster exemplar. Our method first segments the exemplar to extract the primary textons, and then clusters them based on visual similarity. We then compute a descriptor to capture each texton's neighborhood which contains the inter-category relationships that are used at synthesis time. Next, we use a simple procedure to both extract and place the secondary textons behind the primary polygons. Finally, our method constructs a gradient field for the background which is defined by a set of data points and colors. The color of the secondary polygons are also adjusted to better match the gradient field. To compare our work with other methods, we use a wide range of perceptual-based metrics.

Figures (8)

• Figure 1: The high-level pipeline of our texture synthesis algorithm.
• Figure 2: Illustration of descriptors extracted from a texture. Left: primary textons identified in the exemplar. Right: Four sample descriptors. The gray box indicates the boundary of the descriptor, the red polygon is the central texton, and the remaining polygons are the textons included in the descriptor, colored according to category. Notice that several polygons protrude from the initial descriptor boundary.
• Figure 3: The map after different iterations of the synthesis process. White is unknown, light gray is empty, and blue, red, and yellow represent polygons. The newly added descriptor is highlighted with striping.
• Figure 4: Textures synthesized using our algorithm. Each texture pair shows a raster exemplar (left, $500 \times 500$) and a synthetic vector image (right, rendered at $1000 \times 1000$).
• Figure 5: A comparison of our method to Image Quilting ImageQuilting, Self-Tuning Optimization SelfTuningOptimizer, PSGAN PSGAN, and GCD Loss Zhou2023. Default parameters were used.