Saved in:
Bibliographic Details
Main Authors: Zhang, Zeyu, Zhai, Wei, Yang, Jian, Cao, Yang
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2512.18312
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866918257175298048
author Zhang, Zeyu
Zhai, Wei
Yang, Jian
Cao, Yang
author_facet Zhang, Zeyu
Zhai, Wei
Yang, Jian
Cao, Yang
contents The creation of high-fidelity, physically-based rendering (PBR) materials remains a bottleneck in many graphics pipelines, typically requiring specialized equipment and expert-driven post-processing. To democratize this process, we present MatE, a novel method for generating tileable PBR materials from a single image taken under unconstrained, real-world conditions. Given an image and a user-provided mask, MatE first performs coarse rectification using an estimated depth map as a geometric prior, and then employs a dual-branch diffusion model. Leveraging a learned consistency from rotation-aligned and scale-aligned training data, this model further rectify residual distortions from the coarse result and translate it into a complete set of material maps, including albedo, normal, roughness and height. Our framework achieves invariance to the unknown illumination and perspective of the input image, allowing for the recovery of intrinsic material properties from casual captures. Through comprehensive experiments on both synthetic and real-world data, we demonstrate the efficacy and robustness of our approach, enabling users to create realistic materials from real-world image.
format Preprint
id arxiv_https___arxiv_org_abs_2512_18312
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle MatE: Material Extraction from Single-Image via Geometric Prior
Zhang, Zeyu
Zhai, Wei
Yang, Jian
Cao, Yang
Computer Vision and Pattern Recognition
The creation of high-fidelity, physically-based rendering (PBR) materials remains a bottleneck in many graphics pipelines, typically requiring specialized equipment and expert-driven post-processing. To democratize this process, we present MatE, a novel method for generating tileable PBR materials from a single image taken under unconstrained, real-world conditions. Given an image and a user-provided mask, MatE first performs coarse rectification using an estimated depth map as a geometric prior, and then employs a dual-branch diffusion model. Leveraging a learned consistency from rotation-aligned and scale-aligned training data, this model further rectify residual distortions from the coarse result and translate it into a complete set of material maps, including albedo, normal, roughness and height. Our framework achieves invariance to the unknown illumination and perspective of the input image, allowing for the recovery of intrinsic material properties from casual captures. Through comprehensive experiments on both synthetic and real-world data, we demonstrate the efficacy and robustness of our approach, enabling users to create realistic materials from real-world image.
title MatE: Material Extraction from Single-Image via Geometric Prior
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2512.18312