Publication details

Adaptive Measurement of Material Appearance

Thesis

Vávra Radomír


publisher: FIT ČVUT, (Praha 2017)

project(s): GA17-18407S, GA ČR, GA14-02652S, GA ČR, GAP103/11/0335, GA ČR

keywords: material appearance, BRDF, BTF, anisotropy, adaptive measurement, sparse sampling, portable setup, ellipsoidal reflector

preview: Download

abstract (eng):

One of the ultimate challenges of computer graphics is the realistic visualization of appearance of real-world materials. The appearance can be captured by various approaches, but they are often only approximative or usually require an excessively long measurement time. Therefore, this thesis deals with the precise measurement of material appearance utilizing time-reducing adaptive methods. To better understand the behavior of material appearance, we propose an affordable setup for its instantaneous analysis that is based on an ellipsoidal reflector. Also, we study a human's ability to distinguish structure of a material in a virtual environment in dependence on an observation distance. Although, the first proposed method of adaptive measurement does not require a database of already measured materials, it is precise and very flexible. In this approach, the measured space is filled by one-dimensional continuous signals, which are sampled adaptively. We study the optimal deployment of the signals and propose an interpolation method that enables a quick reconstruction of an arbitrary value. Next, we introduce adaptive approaches that rely on the database. Our template-based methods use precomputed sampling patterns for the measurement of a new material and they achieve better results than conventional methods for more than several hundred samples. On the other hand, our minimal-sampling method achieves outstanding results for less than one hundred samples. It is based on the acquisition of a few samples for each rotation of a material around its normal. Therefore, a measurement setup can be very simple or even industrial multi-angle reflectometers can be used. Among adaptive methods, we introduce a non-adaptive image-based approach for acquisition of a huge number of material samples from a large homogeneous specimen. Also, we use gathered knowledge on material appearance to build an inexpensive setup for the rapid acquisition of approximative datasets and propose a novel method for the correct registration of multi-view images. To sum up, our approaches to analysis and measurement have great potential to improve the efficiency of current material appearance acquisition methods.

RIV: BD