Interactive Appearance Manipulation of Fiber-based Materials

Stefan Krumpen, Michael Weinmann and Reinhard Klein

Institute of Computer Science II, University of Bonn, Bonn, Germany

Keywords:

Reﬂectance Modeling, Bidirectional Texture Functions, Interactive Rendering.

Abstract:

Achieving a visually appealing experience for the user interaction with photo-realistic digitized micro-ﬁber

materials is a challenging task. While state-of-the-art high-quality fabric modeling techniques rely on com-

plex micro-geometry representations that are computationally expensive and not well-suited for interactive

rendering, previous interactive reﬂectance models reach a speed-up at the cost of discarding many of the ef-

fects of light exchange that signiﬁcantly contribute to the appearance of fabric materials. In this paper, we

present a novel, example-based technique for the interactive manipulation of micro-ﬁber materials based on

bidirectional texture functions (BTFs) that allow considering ﬁne details in the surface reﬂectance behavior.

BTFs of the respective material sample are acquired for varying ﬁber orientations and combined to a single

texture representation that encodes material appearance depending on the view and light conditions as well

as the orientations of the ﬁbers. This model can be efﬁciently evaluated depending on the user input which,

as demonstrated by our results, allows a realistic simulation of the interaction with micro-ﬁber materials in

real-time.

1 INTRODUCTION

Due to their wide-spread application for e.g. cloth,

towels or furniture, textiles are among the most im-

portant materials we encounter in our daily life. With

the ongoing trend towards the creation of realistic

content for industrial applications in entertainment or

advertisement, there is also an industrial demand for

accurately capturing the characteristics of textiles as

well as modeling the changes in appearance that are

induced by user-manipulations.

When considering textiles, their structural and op-

tical complexity can be seen in the huge number of

individual variants ranging from micro-ﬁber materials

to ﬂuffy carpets. It is not only the reﬂectance behav-

ior of the individual ﬁbers but also the surface struc-

tures at different scales that determine the appearance

of textiles. While larger structures such as the in-

volved yarns inﬂuence the appearance characteristics

of ﬂuffy carpets, there is no such yarn level present

for micro-ﬁber materials and, instead, the orientations

of the small ﬁbers on the surface have a major inﬂu-

ence on the reﬂectance behavior. Unfortunately, ac-

curately modeling textile materials is challenging as

complex, mesoscopic effects of light exchange such

as the self-masking, self-occlusions, scattering within

the ﬁbers and parallax effects induced by the ﬁbers

occur at the surface. State-of-the-art approaches typ-

ically rely on representing the micro-scale geometry

of fabrics in terms of high-resolution volumes, ﬁber

curves or procedural, ﬁber-based models which carry

the burden of high computational demands.

An even more challenging scenario includes the

additional interactive manipulation of digital material

representations. As we all know, moving ﬁngers over

micro-ﬁber materials induces a change in the structure

of the underlying material, i.e. the small ﬁbers are re-

directed according to the orientation of the movement

while certain constraints such as a possibly dominat-

ing ﬁber structure due to the manufacturing process

are met. The re-orientation of the ﬁbers, in turn, leads

to a change in appearance when being touched. In this

paper, we aim at reproducing this painting-like inter-

action for digitized materials.

Unfortunately, the computational effort of the

above-mentioned high-quality fabric modeling tech-

niques renders them impractical for an interactive ma-

nipulation in real-time. To overcome this problem, we

present a novel, example-based technique for the in-

teractive manipulation of micro-ﬁber materials. Our

approach is based on the observations that already

a small number of equilibrium states of ﬁber ori-

entations are sufﬁcient to describe the complex ap-

pearance of a ﬁber-based material and that painting-

like interactions with the material change these states.

This, in turn, changes the appearance in the corre-

266

Krumpen S., Weinmann M. and Klein R.

Interactive Appearance Manipulation of Fiber-based Materials.

DOI: 10.5220/0006168902660273

In Proceedings of the 12th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications (VISIGRAPP 2017), pages 266-273

ISBN: 978-989-758-224-0

sponding area. Furthermore, we argue that direction-

depending reﬂectance characteristics as given by ori-

ented ﬁbers can be rendered photo-realistically using

bidirectional texture functions (BTFs) (Dana et al.,

1997) that capture mesoscopic effects in their data-

driven representation. Therefore, we represent the ap-

pearance of the individual states of the material us-

ing BTFs but other representations such as spatially

varying bidirectional reﬂectance distribution func-

tions (SVBRDFs) or anisotropic microﬂake models

(Jakob et al., 2010) might be used as well. Our

example-based technique allows for the interactive

manipulation of micro-ﬁber materials by combining

the individual BTFs into a single representation that

can be evaluated in real-time and allows a realistic

simulation of the interaction with micro-ﬁber materi-

als. In terms of realism, our approach clearly outper-

forms previous approaches for the interactive manip-

ulation of micro-ﬁber surfaces (Velinov and Hullin,

2016) that rely on SVBRDFs and cannot reproduce

ﬁne details of light exchange induced by the ﬁbers.

2 MICRO-FIBER MATERIALS

With our approach, we focus on micro-ﬁber materi-

als such as velvet, plush, ﬂannel, towels, suede and

alcantara. Some examples are shown in Figure 1.

These materials are characterized by loose micro-

scale ﬁbers of different lengths that are not aligned

with the underlying macroscopic surface geometry.

Depending on the manufacturing process, these ﬁbers

might have certain dominant orientations. These ori-

entations can be inﬂuenced by external forces such

as tactile user interaction. Pressing and moving e.g.

with a ﬁnger over such materials gives the impression

of a painting-like interaction as the ﬁber orientations

are changed. After leaving a certain region with the

ﬁnger, the ﬁbers in this region re-orient to an equilib-

rium state that is responsible for the appearance, i.e.

the reﬂectance behavior of the material in the corre-

sponding area. Depending on the equilibrium state,

the ﬁbers also might occlude each other, cast shadows

and produce interreﬂections. This leads to variations

in material appearance depending on the orientations

of the ﬁbers and the respective view-light conditions.

Figure 2a) shows a material with a predominant

ﬁber orientation in its initial state. In this example,

the dominant ﬁber orientation is not upright but rather

along a certain direction over the surface. User in-

teraction in terms of sweeping the ﬁngers over the

material into or against the dominant ﬁber direction

induces different equilibrium states of the ﬁbers that

inﬂuence the appearance as can be seen in Figure 2b).

Figure 1: Exemplary samples for micro-ﬁber materials.

Figure 2: Illustration of the appearance changes due to dif-

ferent ﬁber orientations: (a) initial state where the ﬁbers

are mainly oriented upright, (b) state where the ﬁbers are

brushed along and against the dominant ﬁber orientation de-

ﬁned by the material structure, and (c) state where the ﬁbers

are brushed perpendicular to the dominant ﬁber orientation

and along the opposite direction.

If the manipulation is carried out perpendicular to

the dominant ﬁber direction, the appearance does not

change if the direction is inverted (see Figure 2c)).

For some materials, additional states have to be taken

into account, e.g. if pressure is applied.

The key observation demonstrated by this exam-

ple is the fact that for a large number of micro-ﬁber

materials it is sufﬁcient to consider only a ﬁnite num-

ber of states. Depending on the material character-

istics, this number might vary but is expected to be

rather small. For example, if the material has no dom-

inant ﬁber orientation, the consideration of the ini-

tial state where the ﬁbers stand upright and the state

where the ﬁbers are brushed along an arbitrary direc-

tion is sufﬁcient. Capturing the states relevant for a

certain material hence allows the interactive synthesis

of the digitized counterpart of the material.

Interactive Appearance Manipulation of Fiber-based Materials

267

3 RELATED WORK

Our approach represents a connection between the

photo-realistic modeling and rendering of fabrics and

the interactive material synthesis. In the following,

we discuss the most related work in these domains.

Appearance Representation for Textile Materials.

Due to the wide-spread use of fabric materials in

graphics applications in visual prototyping, advertise-

ment or entertainment, the realistic appearance mod-

eling of fabrics has gained attention and intensively

investigated in the literature. We only brieﬂy dis-

cuss the most recent developments and refer to re-

spective surveys (Yuen and W

unsche, 2011; Schr

oder

et al., 2012) for more detailed discussions. While

early investigations focused on BRDF-based micro-

facet models (Ashikmin et al., 2000), recent state-of

the-art techniques rely on the modeling of the micro-

scale geometry of fabrics using volumetric scattering

models (Schr

oder et al., 2011; Jakob et al., 2010;

Zhao et al., 2011) and ﬁber-based models (Irawan

and Marschner, 2012; Sadeghi et al., 2013; Khun-

gurn et al., 2015; Schr

oder et al., 2015; Zhao et al.,

2016). The latter approaches can be used in com-

bination with BRDF models, models based on bidi-

rectional ﬁber scattering distribution functions (BFS-

DFs) or models based on bidirectional curve scatter-

ing distribution functions (BCSDFs) for modeling the

reﬂectance behavior of the individual ﬁbers.

The reason for the success of such micro-scale

models lies in the detailed consideration of individ-

ual characteristics such as ﬁber orientations and re-

ﬂectance behavior of the individual ﬁbers which al-

lows to accurately represent the light exchange on

fabric surfaces. However, such high-quality fabric

models have high computational demands as well

as high memory requirements and, hence, cannot be

used for real-time rendering but only for static scenes.

In contrast, the interactive simulation of fabrics re-

quires more light-weight reﬂectance models as e.g.

the adjustment of millions of ﬁbers and the recalcu-

lation of the light exchange are too costly even for

current graphics hardware. With the goal of speed-

ing up the times required for the rendering of fab-

rics, bidirectional texture functions have been synthe-

sized based on known micro-geometry in (Schr

oder

et al., 2013). Since their introduction (Dana et al.,

1997), BTFs have become a popular data-driven re-

ﬂectance model for the photo-realistic depiction of a

huge variety of materials with a surface reﬂectance

behavior ranging from diffuse to glossy to even local

subsurface scattering characteristics. A BTF repre-

sents the reﬂectance behavior at the spatial position

x on the object surface depending on the direction

of the incoming light and the view direction ω

and is therefore deﬁned as a six-dimensional func-

tion ρ

BT F

(x, ω

, ω

). For detailed surveys on BTFs,

we refer to (Haindl and Filip, 2013; Schwartz et al.,

2014; Weinmann et al., 2016). BTFs are parame-

terized on a ﬂat approximation of the true surface.

This allows capturing mesoscopic effects of light ex-

change such as self-occlusions, self-shadowing or in-

terreﬂections that occur in surface scratches of ﬁber-

based materials as well as local subsurface scattering

in the data-driven reﬂectance representation. While

it is well-known that BTFs can be efﬁciently com-

pressed (M

uller, 2009) and used for interactive object

visualization (e.g. (Schwartz et al., 2013a)), the edit-

ing of BTFs is difﬁcult. The latter is the the reason

for the development of other approaches that use sim-

pler reﬂectance models based on SVBRDFs. Most

closely related to the goal of our work is the ﬁtting

of an anisotropic SVBRDF model based on a single-

layer microﬂake model that can be edited interactively

by manipulating its parameters (Velinov and Hullin,

2016). While this technique aims at a small memory

consumption and high rendering speed, this has been

achieved at the cost of not accurately modeling the

aforementioned mesoscopic effects, that contribute

to the appearance of micro-ﬁber materials. While

such ﬁne effects of light exchange are captured in

the image-based BTF, including such effects in an

SVBRDF-based representation would require the ex-

plicit modeling of ﬁne surface details in the geometry.

In contrast, our approach is directly based on

BTFs that are acquired using state-of-the-art acquisi-

tion devices and do not require the ﬁtting of a micro-

ﬁber model. In order to allow an interaction with the

respective material, several BTFs acquired for differ-

ent states of possible ﬁber orientations are combined

into a single model and, during the interactive syn-

thesis, texture lookups are used to achieve a visually

pleasing impression of the material appearance. As

we use BTFs, we have a higher memory footprint but

still achieve interactive rendering performance while

considering the aforementioned effects. Therefore,

our results yield a better visual experience of the ma-

terial characteristics. To the best of our knowledge,

our approach is the ﬁrst interactive framework for ma-

nipulating ﬁber-based materials using BTFs.

Interactions with Texture. Interactions with tex-

ture has become a well-studied topic. The most recent

approaches include the interactive, data-driven high-

ﬁdelity painting system RealBrush (Lu et al., 2013)

and the interactive, example-based texture painting

approach presented in (Luk

c et al., 2015). The lat-

GRAPP 2017 - International Conference on Computer Graphics Theory and Applications

268

ter approach allows to transfer textures with dominant

orientation features onto user-speciﬁed, global direc-

tion ﬁelds while preserving local texture structures.

While in this work only directional features of 2D tex-

tures are exploited to generate new textures that fol-

low a user-speciﬁed direction ﬁeld, our technique al-

lows to interactively modify the direction ﬁeld and at

the same time synthesize novel textures from a set of

pre-deﬁned textures that represent different equilib-

rium states. This way, it allows for the manipulation

of complex surface reﬂectance characteristics.

Time-varying or Interactive Material Synthesis.

Many applications consider time-varying material

appearance. Seminal work focused on modeling

changes in appearance for burning, drying, corrosion

and decaying processes and proposed the ﬁrst densely

measured database of time and spatially-varying ap-

pearance of ﬂat samples (Gu et al., 2006). Similar

effects were considered in (Sun et al., 2006) with the

drying of different paints or wet surfaces such as ce-

ment, plaster and fabrics, and the accumulation of

dusts on surfaces and the melting of materials such as

chocolate. In both works, ﬂat materials were consid-

ered and analytical BRDFs were ﬁt to acquired mea-

surements performed under varying view-light con-

ditions and at multiple time instances. Furthermore,

time-varying BTFs have been proposed that also con-

sider mesoscopic effects (Langenbucher et al., 2010).

Interactive material design has been studied by

considering manipulations on both small-scale geom-

etry and materials (Wu et al., 2011). However, scat-

tering effects that are important for accurate appear-

ance reproduction at small-scale structures as well

as anisotropic material characteristics are not consid-

ered. More recently, variations in material appearance

induced by interactive user manipulation have been

modeled based on an anisotropic SVBRDF model

(Velinov and Hullin, 2016).

4 METHODOLOGY

In this section, we ﬁrst provide an overview of the

proposed technique and subsequently discuss the re-

spective key components in more detail.

The main components of our techniques are il-

lustrated in Figure 3. In a pre-processing stage, we

acquire micro-ﬁber materials in different states such

as different ﬁber orientations (see Section 2) using

a state-of-the-art reﬂectance acquisition device. This

results in the computation of one BTF for each equi-

librium state of the ﬁbers. The resulting BTFs are

then combined into a single texture array that holds

the information regarding the appearance of a respec-

tive material under different view-light conﬁgurations

and different orientations of the ﬁbers. In the ﬁnal on-

line stage, an interactive appearance simulation takes

the per-material texture array to specify an initial con-

ﬁguration of view and light conditions as well as the

ﬁber directions and allows the user to virtually brush

the material with his/her ﬁngers and change the view

and light conditions, while the appearance changes

accordingly.

4.1 Reﬂectance Representation and

Acquisition

As the reproduction of ﬁne details in the reﬂectance

behavior induced by the individual ﬁbers is of utmost

importance for a visually appealing impression of the

material, we focus on the use of BTFs for modeling

the reﬂectance behavior of micro-ﬁber materials. In

particular, we measure an individual material sample

several times for different ﬁber states τ with a stan-

dard acquisition device (Schwartz et al., 2013b). A

material sample with a certain ﬁber state τ is placed

onto a turntable and observed by eleven cameras that

are mounted on an arc with increments of 7.5

◦

and

have a resolution of 2,048 × 2,048 pixels. After an

initial structured light based acquisition of the approx-

imate surface, the 198 LED light sources mounted on

the upper hemisphere are used to sequentially illumi-

nate the material sample from different light source

positions while the material sample is captured by

the cameras under different turntable rotations. Af-

ter the acquisition, we project the captured HDR im-

ages onto the surface which is followed by a resam-

pling of the data into local coordinate systems at each

surface point and a ﬁnal Decorrelated Full-Matrix-

Factorization (DFMF) compression. During the com-

pression, an SVD is computed for each color channel

after a conversion into the YUV color space, which

results in Eigen-ABRDFs and Eigen-Textures. From

these, the most informative ones are kept to achieve

the compression. Finally, this results in an individual

BTF for each of the ﬁber states. In order to obtain a

single reﬂectance representation that also encodes the

changes in reﬂectance behavior induced by the ﬁber

characteristics in addition to the view and light di-

rections, these BTFs ρ

BT F

(x, ω

, ω

) are merged into

a single BTF ρ

BT F

(x, ω

, ω

, τ) that additionally in-

cludes the dependency on the ﬁber state τ.

4.2 Interactive Simulation

The data-driven representation resulting from the pre-

vious step models material appearance depending on

Interactive Appearance Manipulation of Fiber-based Materials

269

merged

BTF

BTF 1

BTF 2

Appearance Acquisition

Samples with varying

fiber orientations

Appearance Modeling

User Input

Interactive Simulation

Figure 3: Overview of the main components of the proposed technique: In an initial step, micro-ﬁber materials with different

ﬁber orientations are acquired with a reﬂectance acquisition device which results in one BTF per ﬁber orientation of the

material. The individual BTFs are combined into a single texture array that contains the reﬂectance characteristics of the

material under varying view and light conditions as well as varying ﬁber orientations. Based on this representation and an

input by the user, an interactive simulation of the reﬂectance behavior for manipulated ﬁber orientations is achieved.

the spatial position on the micro-ﬁber surface, the

view direction, the light direction and the state τ of the

ﬁbers. Before we describe how an efﬁcient rendering

can be performed despite the rather high dimensional-

ity of the measure data, we ﬁrst discuss the realization

of user interactions with micro-ﬁber materials.

User Interaction. To achieve an appealing experi-

ence for the painting-like interaction of the user with

the material, we aim at an as-intuitive-as-possible in-

terface that allows the user to brush over the digi-

tized micro-ﬁber material. In more detail, the mate-

rial is shown to the user who can change the view-

ing and lighting conditions and is allowed to directly

paint on the object via the mouse. The latter is imple-

mented by projecting the mouse position into the uv-

parametrization of the object and using the resulting

coordinates to draw in a direction texture. This tex-

ture has three color channels that hold the direction of

the stroke (d

, d

) in uv-space in the red and green

color channels, and a brush-strength s in the blue-

channel, where s = 1 means that the material is com-

pletely brushed and s = 0 means that the correspond-

ing equilibrium state remains unchanged. To take into

account that a hard transition between brushed and

unbrushed ﬁbers at the edges of a stroke is unrealistic

for the simulated materials, we interpolate s between

0 and 1 at the edges of the stroke. By interpolating

the corresponding BTFs of the borders of the brush-

stroke according to s, we approximate the reﬂectance

behavior at these parts. Please note, that a more ac-

curate interpolation might be achieved based on more

sophisticated techniques (Bonneel et al., 2011), how-

ever, simple linear interpolation allows for faster ren-

dering. While drawing, we brush all ﬁbers below the

virtual ”‘ﬁnger”’ in the direction of the movement of

the ﬁnger independent of their position relative to the

brush center, but according to the resulting equilib-

rium state. This state is chosen depending on the an-

gle between the movement direction of the ﬁnger in

texture space and the dominant ﬁber direction stored

with each material. At each frame, only the parts of

the direction texture which were changed during the

frame are updated in the video memory.

Rendering. Based on the aforementioned direction

texture generated by the user interactions, the re-

ﬂectance model from Section 4.1 has to be efﬁciently

evaluated on the GPU. For this purpose, the measured

BTFs, consisting of the Eigen-Textures and Eigen-

ABRDFs for the three color-channels, and the direc-

tion texture containing the user generated direction

ﬁeld and brush-strength information are passed to the

fragment shader. The shader samples the direction-

texture to determine which state of the material has

to be applied for the current fragment, and how the

corresponding BTF has to be rotated in order to align

its orientation according to the user-deﬁned direction

texture. Note, that the BTFs are oriented in such a

way that the positive u-axis of the corresponding 2D

texture is aligned with its dominant ﬁber direction. If

the brush strength is s = 0, the shader only evaluates

the BTF representing the material in its initial state.

For s = 1, the BTF for the state τ corresponding to the

direction of the brush stroke is evaluated. If 0 < s < 1,

which occurs on the edges of a brush stroke, the BTFs

acquired for the initial state and the state with the ad-

equate ﬁber orientations are evaluated and the results

are interpolated. The material state which needs to be

sampled is determined from the angle α between the

GRAPP 2017 - International Conference on Computer Graphics Theory and Applications

270

vector (1, 0) in uv-space and the direction d = (d

, d

)

of the brush stroke. Since only a few ﬁber orienta-

tions for a few angles were measured, we select the

one which is closest to the angle α of the brush stroke.

The directions ω

and ω

used to sample the BTF are

deﬁned w.r.t. the local coordinate system at x. For

sampling the material brushed into a certain direction

d at a texel x, we rotate the tangent t deﬁning the

local coordinate system by −α = −atan(

) around

the normal n, and the texture-coordinate by α around

the center of the texture. The rotated coordinate sys-

tem and texture-coordinate are then used to sample

the BTF of the brushed material.

5 RESULTS

We evaluate our technique with respect to its capabil-

ity to accurately reproduce the characteristic material

appearance during user interaction and its rendering

performance.

5.1 Visual Quality

To evaluate the proposed method regarding the

achieved visual quality, we measured BTFs of a ﬁber

based material sample in two different states: One

state where all ﬁbers of the material are arranged

upwards, and a second state where all ﬁbers were

brushed into one direction. As the used material does

not have a predominant direction, it is sufﬁcient to

capture only one direction of the ﬁbers, as the ap-

pearance is invariant to the direction of interaction.

Figure 4 shows original images of the material sam-

ple, lit from different directions, were the ﬁbers in

the middle were brushed from right to left. It is

clearly visible that the individual parts have a dif-

ferent reﬂectance behavior when the light direction

changes. Figure 5 shows how the material rendered

using our method behaves when the light or view di-

rection is changed. Mesoscopic effects of light ex-

change such as interreﬂections, self-shadowing and

self-occlusions are preserved in the synthesized mate-

rial as well as parallax effects. As the used BTFs are

not tileable, there are discontinuities in the textures.

Figure 6 shows a comparison to a previous technique

(Velinov and Hullin, 2016). Furthermore, Figure 7

shows a rendering of the material applied to a curved

object using an environment map.

5.2 Rendering Performance

All of our renderings were performed on a machine

with a NVidia Geforce 980 GTX with 4GB of VRAM.

The most costly part of our shader is the evaluation

of the BTFs, as this comes with a high amount of

texture lookups per fragment, especially when both

BTFs have to be evaluated on the edges of a brush-

stroke. The visual quality of BTFs greatly depends

on the number of BTF components used for ren-

dering. In our evaluations, we obtained the insight

that using 50 Eigen-Textures and Eigen-ABRDFs for

the brightness-channel and eight Eigen-Textures and

Eigen-ABRDFs for the color channels is sufﬁcient for

a good visual quality, while allowing for high framer-

ates. For rendering without user manipulations, we

achieved framerates of 125 frames per second in av-

erage with the BTF settings mentioned above for the

scene in the bottom row of Figure 5. During user

interaction, where parts of the direction texture are

changed and uploaded to the GPU, which is a rather

slow operation, the framerate drops down to 80 FPS.

The used BTFs have a resolution of 900 × 400 pixels,

and about 50 MB of video memory each for the used

settings.

6 CONCLUSION

In this paper, we proposed a data-driven method to

interactively manipulate ﬁber-based materials. As

demonstrated in our experiments, our technique is

capable of reproducing several effects such as self-

occlusions, self-shadowing or the scattering between

the ﬁbers during the interactive simulation and,

hence, clearly outperforms the previous state-of-the-

art (Velinov and Hullin, 2016) that cannot reproduce

such effects reliably in terms of realism. To overcome

the limited size of the measured BTFs, one could ap-

ply the concept of video-textures (Sch

odl et al., 2000)

to hide the tiling-artifacts. This could be achieved by

searching for patches inside the BTF for the current

material state that are similar to the one currently un-

der the brush, instead of moving over the border of

the BTF texture. Our framework relies on the avail-

ability of BTFs measured for different material states.

For certain materials and scenarios, such as brushing

ﬁber-based materials, where only a few states have to

be considered, this simple approach is is highly ef-

ﬁcient. To include further interactions such as com-

pacting the ﬁbers, additional states of the considered

material have to be measured. This means that the

total acquisition effort might become impractical if

many states have to be considered. However, as at

Interactive Appearance Manipulation of Fiber-based Materials

271

Figure 4: The material sample used for our experiments under different lighting conditions where parts of the sample were

brushed into the direction indicated by the red arrow using a ﬁnger. Note that the ﬁbers in the unbrushed parts are not arranged

straight upwards, which is why there is a larger change in brightness also in these parts.

Figure 5: Renderings of the captured material using our technique: When brushing the ﬁbers into different directions (top

row), the characteristic differences in appearance induced by the ﬁber orientations are clearly visible as the light direction

changes. Furthermore, renderings of two ﬁnger-strokes obtained using our technique are shown (bottom row). The arrows

in the bottom-left image indicate the directions in which the virtual ﬁnger was moved. The white sphere is included to

better visualize the light direction. The ﬁbers in the brushed areas behave differently depending on their orientation and the

view-light conditions. This ﬁgure is best viewed in color and by using the zoom function.

Figure 6: Comparison between the real material (upper

left), our measured BTF (upper right) and renderings of

the manipulated material obtained using a state-of-the-art

technique (Velinov and Hullin, 2016) (bottom left) and our

technique (bottom right). Please note that the lighting con-

ditions are different in the photo and the renderings.

Figure 7: Rendering of the material with several brush-

strokes under environment-lighting. The environment is ap-

proximated by eight directional lights.

most two BTFs have to be evaluated for a single frag-

ment, the efﬁciency of the rendering pipeline remains

unchanged but the memory consumption increases.

Furthermore, a more general editing of the materi-

als such as changing its colors is not possible. Such

a material editing might be implemented by ﬁtting

GRAPP 2017 - International Conference on Computer Graphics Theory and Applications

272

SVBRDFs to the BTF data, storing the residual of the

Y channel in a BTF and adding the residual again dur-

ing the rendering process.

REFERENCES

Ashikmin, M., Premo

ze, S., and Shirley, P. (2000). A

microfacet-based BRDF generator. In Proceedings of

the 27th Annual Conference on Computer Graphics

and Interactive Techniques, pages 65–74.

Bonneel, N., van de Panne, M., Paris, S., and Heidrich, W.

(2011). Displacement interpolation using lagrangian

mass transport. ACM Trans. Graph., 30(6):158:1–

158:12.

Dana, K. J., Nayar, S. K., van Ginneken, B., and Koen-

derink, J. J. (1997). Reﬂectance and texture of real-

world surfaces. In Proceedings of the IEEE Con-

ference on Computer Vision and Pattern Recognition

(CVPR), pages 151–157.

Gu, J., Tu, C.-I., Ramamoorthi, R., Belhumeur, P., Matusik,

W., and Nayar, S. (2006). Time-varying surface ap-

pearance: Acquisition, modeling and rendering. ACM

Trans. Graph., 25(3):762–771.

Haindl, M. and Filip, J. (2013). Visual Texture: Accu-

rate Material Appearance Measurement, Representa-

tion and Modeling. Advances in Computer Vision

and Pattern Recognition. Springer-Verlag New York

Incorporated.

Irawan, P. and Marschner, S. (2012). Specular reﬂection

from woven cloth. ACM Trans. Graph., 31(1):11:1–

11:20.

Jakob, W., Arbree, A., Moon, J. T., Bala, K., and

Marschner, S. (2010). A radiative transfer frame-

work for rendering materials with anisotropic struc-

ture. ACM Trans. Graph., 29(4):53:1–53:13.

Khungurn, P., Schroeder, D., Zhao, S., Bala, K., and

Marschner, S. (2015). Matching real fabrics with

micro-appearance models. ACM Trans. Graph.,

35(1):1:1–1:26.

Langenbucher, T., Merzbach, S., M

oller, D., Ochmann, S.,

Vock, R., Warnecke, W., and Zschippig, M. (2010).

Time-varying BTFs. In Central European Seminar on

Computer Graphics for Students (CESCG).

Lu, J., Barnes, C., DiVerdi, S., and Finkelstein, A. (2013).

Realbrush: Painting with examples of physical media.

ACM Trans. Graph., 32(4):117:1–117:12.

Luk

c, M., Fi

ser, J., Asente, P., Lu, J., Shechtman, E., and

ykora, D. (2015). Brushables: Example-based edge-

aware directional texture painting. Comput. Graph.

Forum, 34(7):257–267.

uller, G. (2009). Data-Driven Methods for Compression

and Editing of Spatially Varying Appearance. Disser-

tation, Universit

at Bonn.

Sadeghi, I., Bisker, O., De Deken, J., and Jensen, H. W.

(2013). A practical microcylinder appearance model

for cloth rendering. ACM Trans. Graph., 32(2):14:1–

14:12.

Sch

odl, A., Szeliski, R., Salesin, D. H., and Essa, I. (2000).

Video textures. In Proceedings of the Annual Con-

ference on Computer Graphics and Interactive Tech-

niques, SIGGRAPH ’00, pages 489–498.

Schr

oder, K., Klein, R., and Zinke, A. (2011). A volumetric

approach to predictive rendering of fabrics. In Pro-

ceedings of the Eurographics Conference on Render-

ing (EGSR), pages 1277–1286.

Schr

oder, K., Klein, R., and Zinke, A. (2013). Non-

local image reconstruction for efﬁcient computation

of synthetic bidirectional texture functions. Computer

Graphics Forum, 32:61–71.

Schr

oder, K., Zhao, S., and Zinke, A. (2012). Recent ad-

vances in physically-based appearance modeling of

cloth. In SIGGRAPH Asia 2012 Courses, pages 12:1–

12:52.

Schr

oder, K., Zinke, A., and Klein, R. (2015). Image-based

reverse engineering and visual prototyping of woven

cloth. IEEE Transactions on Visualization and Com-

puter Graphics, 21(2):188–200.

Schwartz, C., Ruiters, R., Weinmann, M., and Klein, R.

(2013a). Webgl-based streaming and presentation of

objects with bidirectional texture functions. J. Com-

put. Cult. Herit., 6(3):11:1–11:21.

Schwartz, C., Sarlette, R., Weinmann, M., and Klein, R.

(2013b). DOME II: A parallelized BTF acquisition

system. In Proceedings of the Eurographics Workshop

on Material Appearance Modeling, pages 25–31.

Schwartz, C., Sarlette, R., Weinmann, M., Rump, M., and

Klein, R. (2014). Design and implementation of prac-

tical bidirectional texture function measurement de-

vices focusing on the developments at the University

of Bonn. Sensors, 14(5):7753–7819.

Sun, B., Sunkavalli, K., Ramamoorthi, R., Belhumeur, P.,

and Nayar, S. (2006). Time-varying BRDFs. In Pro-

ceedings of the Second Eurographics Conference on

Natural Phenomena (NPH), pages 15–23.

Velinov, Z. and Hullin, M. B. (2016). An Interactive Ap-

pearance Model for Microscopic Fiber Surfaces. In

Hullin, M., Stamminger, M., and Weinkauf, T., edi-

tors, Vision, Modeling and Visualization.

Weinmann, M., Langguth, F., Goesele, M., and Klein, R.

(2016). Advances in geometry and reﬂectance acqui-

sition. In Eurographics 2016 Tutorials.

Wu, H., Dorsey, J., and Rushmeier, H. (2011). Physically-

based interactive bi-scale material design. In Proceed-

ings of the 2011 SIGGRAPH Asia Conference, pages

145:1–145:10.

Yuen, W. and W

unsche, B. C. (2011). An evaluation on

woven cloth rendering techniques. In Proceedings of

the International Image and Vision Computing New

Zealand Conference (IVCNZ 2011), pages 7–12.

Zhao, S., Jakob, W., Marschner, S., and Bala, K. (2011).

Building volumetric appearance models of fabric

using micro CT imaging. ACM Trans. Graph.,

30(4):44:1–44:10.

Zhao, S., Luan, F., and Bala, K. (2016). Fitting procedural

yarn models for realistic cloth rendering. ACM Trans.

Graph., 35(4):51:1–51:11.

Interactive Appearance Manipulation of Fiber-based Materials

273