[Paper Review] Computational Parquetry: Fabricated Style Transfer with Wood Pixels

Julian Iseringhausen, Michael Weinmann+2 more|arXiv (Cornell University)|Apr 9, 2019

Computer Graphics and Visualization Techniques64 references4 citations

TL;DR

This paper introduces a computational pipeline for fabricating parquetry art by using real wood veneers with natural grain and color variations as 'pixels' to reconstruct target images. The method employs a combinatorial optimization approach to assign unique wood patches to image regions while avoiding collisions, enabling high-fidelity, structure-aware wooden artworks using only laser cutting and scanning hardware.

ABSTRACT

Parquetry is the art and craft of decorating a surface with a pattern of differently colored veneers of wood, stone or other materials. Traditionally, the process of designing and making parquetry has been driven by color, using the texture found in real wood only for stylization or as a decorative effect. Here, we introduce a computational pipeline that draws from the rich natural structure of strongly textured real-world veneers as a source of detail in order to approximate a target image as faithfully as possible using a manageable number of parts. This challenge is closely related to the established problems of patch-based image synthesis and stylization in some ways, but fundamentally different in others. Most importantly, the limited availability of resources (any piece of wood can only be used once) turns the relatively simple problem of finding the right piece for the target location into the combinatorial problem of finding optimal parts while avoiding resource collisions. We introduce an algorithm that allows to efficiently solve an approximation to the problem. It further addresses challenges like gamut mapping, feature characterization and the search for fabricable cuts. We demonstrate the effectiveness of the system by fabricating a selection of "photo-realistic" pieces of parquetry from different kinds of unstained wood veneer.

Motivation & Objective

• To develop a method that uses the natural structural details of real wood veneers as a source of visual detail for image reconstruction in physical parquetry.
• To address the challenge of limited, non-reusable wood pieces by formulating a resource-constrained assignment problem that avoids collisions between patches.
• To create a fully computational, end-to-end pipeline that enables non-experts to generate fabricated parquetry puzzles using only a laser cutter and flatbed scanner.
• To preserve the full visual complexity of wood—such as grain, knots, and color variation—by treating each veneer piece as a unique, non-replicable texture source.
• To demonstrate the feasibility of producing high-contrast, shaded, and detailed wooden artworks without relying on color-based simplification or post-processing.

Proposed method

• The system uses a patch-based image synthesis approach adapted for physical fabrication, where each wood veneer patch is treated as a unique, non-reusable image patch.
• A combinatorial optimization algorithm assigns the best-fitting wood patches to target image regions while tracking resource usage to prevent overlaps and ensure each piece is used only once.
• Feature characterization is performed on scanned wood samples to extract color and texture features, enabling accurate similarity matching between source patches and target image regions.
• Gamut mapping is applied to align the color and texture ranges of wood samples with the target image's appearance, improving visual fidelity.
• The method incorporates fabricability constraints by generating cuts that are feasible for laser cutting, including considerations for edge alignment and piece shape.
• A U-Net-based model is trained to predict the visual change from unfinished to finished wood, enabling pre-scanning with surface finish applied to improve appearance prediction accuracy.

Experimental results

Research questions

• RQ1 How can natural structural details in real wood veneers be leveraged as a source of visual detail for physical image reconstruction in parquetry?
• RQ2 What optimization strategy enables high-fidelity image approximation using only unique, non-reusable wood patches without overlap or deformation?
• RQ3 How can the appearance of wood be accurately modeled through scanning when surface finish significantly alters its visual properties?
• RQ4 To what extent can a computational pipeline generate physically fabricatable parquetry puzzles using only commodity hardware and minimal user input?
• RQ5 Can the method produce visually compelling results without relying on color-based simplification or post-processing, preserving the full complexity of wood texture?

Key findings

• The system successfully produced high-contrast, detailed parquetry artworks using only six types of unstained wood veneers and a target image, achieving visually faithful reconstructions.
• The use of natural wood structures—such as grain, knots, and color variation—enabled fine structural details and proper shading, avoiding the 'posterized' look common in traditional methods.
• The combinatorial optimization approach effectively managed resource constraints, ensuring no two patches overlapped while maintaining visual fidelity.
• The U-Net model trained on pre- and post-finish scans showed promising results in predicting appearance changes, suggesting potential for future automation of the finishing step.
• The pipeline enabled non-experts to create complex, visually rich parquetry puzzles using only a laser cutter and flatbed scanner, demonstrating accessibility and practicality.
• The resulting parquetry puzzles featured irregularly shaped pieces that provided strong visual and tactile cues, significantly improving assembly over standard uniform-shaped puzzles.