A Study of Opacity Ranges for Transparent Overlays in 3D Landscapes
AA Study of Opacity Ranges for Transparent Overlays in 3D Landscapes
Jan Hombeck * University of KoblenzUniversity of Victoria
Li Ji † LlamaZOO Interactive Inc.
Kai Lawonn ‡ University of Jena
Charles Perin § University of VictoriaFigure 1: Transparent overlays in a photorealistic 3D environment using striped, filled and dotted internal patterns with an outline. A BSTRACT
When visualizing data in a realistically rendered 3D virtual environ-ment, it is often important to represent not only the 3D scene but alsooverlaid information about additional, abstract data. These overlaysmust be usefully visible, i.e. be readable enough to convey the infor-mation they represent, but remain unobtrusive to avoid cluttering theview. We take a step toward establishing guidelines for designingsuch overlays by studying the relationship between three differentpatterns (filled, striped and dotted patterns), two pattern densities,the presence or not of a solid outline, two types of background (blankand with trees), and the opacity of the overlay. For each combinationof factors, participants set the faintest and the strongest acceptableopacity values. Results from this first study suggest that i) rangesof acceptable opacities are around 20-70%, that ii) ranges can beextended by 5% by using an outline, and that iii) ranges shift basedon features like pattern and density.
Index Terms:
Overlays—Opacity—Visualization
NTRODUCTION AND B ACKGROUND
We investigate which opacity ranges to use when overlaying shapesfor visualizing abstract data within a spatially bounded region of in-terest (RoI) on a photorealistic landscape. Our research is motivatedby the recent adoption of real-time and high-performance photoreal-istic rendering in industrial data visualization applications, and in par-ticular, high-fidelity GIS-based visualization applications [1, 2, 25].The standard way of representing information within a RoI is to over-lay a transparent layer of color on top of the realistically renderedlandscape to delineate a region of interest [9]. This technique is usedfor example in map-based Geographic Information System (GIS)tools such as ESRI ArcGIS [2]. Creating an overlaid visualizationthat sits on top of a photorealistic, 3D environment, such as theexample shown in Figure 1 requires a visualization designer to make * e-mail: [email protected] † e-mail: [email protected] ‡ e-mail: [email protected] § e-mail: [email protected] choices. These choices are currently limited, as they are inheritedfrom 2D GIS tools and constrained to a transparent layer of colorbeing displayed on top of the realistically rendered landscape [9].To think beyond this conventional design we can turn to the fieldof information visualization, which provides many empirical guide-lines for representing abstract data, established through graphicalperception studies [6]. Opacity is an effective visual variable to en-code quantitative information [24] such as uncertainty [7] of the RoI,and has been the focus of studies on the perception of grids in 2Dcharts [5]. Graphical Perception studies in information visualization,however, make strong assumptions regarding the background andenvironment in which the abstract data is being represented, typi-cally a uniform background or a 2D map and a static point of view.In contrast, 3D landscapes are colorful, contain complex geometriesand textures, and can be looked at from infinite numbers of pointsof view. It makes it difficult to directly apply results from studiesthat consider standard 2D charts (e.g., [5, 13, 14, 20]). TUDY OF O PACITY R ANGES FOR O VERLAYS
To identify reasonable opacity ranges for overlays, we must considerthe many other graphical properties of these overlays. Our first stepwas to identify promising ones to study.In the field of illustrative rendering, patterns that are filled [12],striped [15,26] and doted [15,21] are widely used to render overlays.Outlines [17] are commonly used to increase the visibility of3D shapes [18, 19]. Pattern density [10] and overlay opacity [5]also affect how information is perceived. With transparent overlays,the rendered image must contain sufficient visual cue to permitthe Human Visual System (HVS) to recognize the separate layers,instead of a single layer of the blended colors [11, 23].A sketching session with our research group and an interview witha visualization practitioner confirmed the usefulness of investigatingthe following rendering parameters, widely used in the literature:pattern style, pattern density, pattern outline, and scene background.
We conducted a remote study in accordance to our research ethicsboard regulations during the covid-19 pandemic. A remote studymeans lower internal validity of the results [22] because the appara-tus is less controlled than in a lab setting (e.g., varying display sizes,luminosity and contrast). To control for these factors we requiredparticipants to use a monitor with resolution of 1920x1080 px and a r X i v : . [ c s . G R ] S e p U T L I N E PATTERNDENSITY P Fill D Low D High D Low D High P Dots P Stripes O On O Ø Figure 2: The three
PAT T ERN for the two
OUT LINE for the two
DENSITY used in the study (only
BKG is not shown in this Figure). with the luminosity value of their screen set to its maximum, and weconducted the study during daylight hours only. On the other hand,a remote study increases the external validity of the results [22]. Itis closer to the real-world context, where people use various devicesand screens in various environments. Overall, our study loses someprecision to increase its realism and generalizability [22].We conducted the study with the TeamViewer remote desktopcontrol software [3]. This allowed the experimenter to verbally com-municate with the participants, answer their questions and observetheir interactions. Participants interacted with the study software viamouse and keyboard. While we required a good internet connectionfrom participants, the delay (time between the action from the partic-ipant and the response from the server) varied between participantsand sometimes within a session. The study software showed viewsof a forest area in a professionally used virtual environment in Unity.
To establish opacity ranges in which different configurations ofoverlay are considered useful, we created the two following tasksbased on Bartram et al.’s [5] study of opacity values for grids:• T Faintest : Please adjust the color to be as faint as you thinkit can comfortably be to be still useful; any fainter and youwould no longer be able to easily use it. • T Strongest : Please adjust the color to be as strongly visible asyou think it can comfortably be before it interferes with or“comes in front of” the environment; any stronger and it wouldbe too obtrusive.
Participant used the left and right arrow keys to change the opacityvalue and pressed the space bar to confirm and go to the next trial.
We studied the four following factors (see Figures 2 and 3):•
PAT T ERN was either P Fill (a solid color pattern), P Stripes (astripped pattern). or P Dots (a dotted pattern).• Both P Stripes and P Dots were presented with two different val-ues for
DENSITY . With P Stripes at low density D Low , eachstripe as a 240-pixel width and a 120-pixel width at high den-sity D High . With P Dots each dot has a 240-pixel diameter at D Low and a 120-pixel diameter at D High .• The overlay
OUT LINE (a 12-pixel width solid stroke aroundthe RoI) was either present ( O On ) or not ( O ∅ ).• The scene BKG was either with trees B Trees or without B ∅ .Overall the experiment consisted of [ × PAT T ERN ( P Stripes , P Dots ) × × DENSITY ( D Low , D High ) + × PAT T ERN ( P Fill ) ] × × OUT LINE ( O On , O ∅ ) × × BKG ( B ∅ , B Trees ) × × TASK ( T Faintest , T Strongest ) × × T RIAL = 160 TRIALS per participant.Each participant performed the experiment in two task blocks, onefor T Faintest and one for T Strongest (80 trials per task block in randomorder). The order of task block was balanced across participants.
Figure 3: Side by side view of two configurations used during thestudy. P Stripes , D High , O On with B ∅ (left) and B Trees (right). The overlayis rendered in a post-processing step that maps its shape on top ofthe underlying surface (that includes objects present in the scene).
The experimenter introduced the study and let the participant readthe instructions explaining the tasks. The participant then performeda series of training trials until they understood the tasks and controls.After having asked any question they had, they started the recordedtrials, either with T Faintest or with T Strongest first. Once they hadcompleted the first task block, they were asked to take a 5-minutebreak before completing the second task block. Finally they filledout the study questionnaire and participated in a short debriefing.The questionnaire asked demographic questions (age, gender, levelof education, experience in visualization, experience in 3D, usageof video games and of computers) and study-related questions onLikert scales. Each session lasted approximately one hour.
We hypothesize that the denser a pattern, the easier to perceive theoverlay as a conceptually separate layer for the HSV, and thus theless saliency the overlay will need. For instance, T Faintest will leadto higher opacity values for D Low , because a low-density patternwill require more emphasis through higher opacity, for the HVS torecognize the overlay as a conceptually separate layer.We also hypothesize that the outline will facilitate the HVS tomore easily recognize that the overlay belongs to a separate shape,and therefore belongs to a conceptually separate transparent layer.Therefore, the outline will make a wider range of opacity recogniz-able as transparent overlays instead of a single layer of blended color.More formally, our hypotheses were as follows:• H DensityFaint : D Low will lead to higher average opacity valuesthan D High for T Faintest . Considering that P Fill has the highestdensity, we therefore hypothesize that for T Faintest , O( P Fill ) < O( P Stripes × D High ) < O( P Dots × D High ) < O( P Stripes × D Low ) < O( P Dots × D Low ). Indeed, with D Low , the pattern shapes arelarger and there are fewer of them, resulting in relatively largeempty spaces in the pattern.• H DensityStrong : D Low will lead to higher average opacity valuesthan D High for T Strongest . With P Fill having infinite density,we hypothesize that for T Strongest , we will obtain O( P Fill ) < O( P Stripes × D High ) < O( P Dots × D High ) < O( P Stripes × D Low ) < O( P Dots × D Low ). Indeed, we hypothesize that larger emptyspaces make it easier to mentally reconstruct the underlying3D scene thus tolerate greater opacity values.• H Outline : Outlines will enlarge opacity ranges, i.e. for T Faintest ,O( O On ) < O( O ∅ ) and for T Strongest , O( O ∅ ) < O( O On ), be-cause the shape of the overlay is clearer with an outline.• H Background : Trees in the background will lead to smalleropacity ranges, i.e. for T Faintest , O( B ∅ ) < O( B Trees ), becauseadditional content such as trees distort the overlay; and for
TRONGESTOPACITY
Fully transparent Fully opaque Faster thanaverage Slower thanaverage Faster thanaverage Slower thanaverage
SECONDS SECONDS
OPACITY VALUE DEVIATION FROM PATICIPANT AVERAGE COMPLETION TIME
FAINTEST STRONGESTFAINTEST P A T T E R N D E N S I T Y O U T L I N E O U T L I N E O N O U T L I N E O FF B A C K G R O U N D W I T H T R EE S W I T H O U T T R EE S O U T L I N E O N O U T L I N E O FF DOTS HIGHDOTS LOWDOTS HIGHDOTS LOWDOTS HIGHDOTS LOWDOTS HIGHDOTS LOWSTRIPES HIGHSTRIPES LOWSTRIPES HIGHSTRIPES LOWSTRIPES HIGHSTRIPES LOWSTRIPES HIGHSTRIPES LOWFILLFILLFILLFILL 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 -6 -4 -2 0 +2 +4 +6 -6 -4 -2 0 +2 +4 +60.0
Correspondence between degree of confidence interval overlap and p-values for independent samples
95% CIs p -value .1.001.005.01.05 Figure 4: Opacity values and deviation from average completion time mean 95% bootstrappd CIs for each task, grouped by
BKG , then
OUT LINE ,then
PAT T ERN and then
DENSITY . The correspondence between degree of CI overlap and p-values for independent samples is based on [16].The rectangles between mean estimates for opacity values give an indication of the width of the opacity range for each condition. T Strongest , O( B Trees ) < O( B ∅ ), because comfortably seeing thetrees might require lower opacity values. We recruited sixteen participants (six male, ten female) with nor-mal or corrected vision through mailing lists and social networks.Most participants were between 19–25 years of age with on average1-2 years of experience with 3D environments and data visualiza-tion. Detailed participant information is provided in supplementalmaterial. Participants received $15 via email for their participation.
We base our analyses on estimation using bootstrapped confidenceintervals instead (CI) of p-values, following recommendations fromAPA [4]. A 95% CI contains the true mean 95% of the time andconveys effect sizes [8]. We pre-specified all analyses before run-ning the study and tested on pilot data (scripts for parsing the data,computing CIs, and generating drafts of figures).
Here we revisit our hypotheses against the results shown in Figure 4. H DensityFaint is partially confirmed. Indeed, P Fill has consistentlylower opacity values than the other configurations of
PAT T ERN and
DENSITY . However, P Stripes × D Low has a significantly loweropacity value than P Stripes × D High ; and P Dots × D Low has a signif-icantly lower opacity value than P Dots × D High – these results areopposite to what we hypothesized. H DensityStrong is not confirmed,with only P Fill and P Stripes × D High leading weakly to greater opac-ity values than P Dots × D High (around 5% greater). Overall, lowerdensity leads to lower values for the lower bound of the opacityrange, resulting in larger usable opacity ranges (around 5% largerfor P Stripes and 10% larger for P Dots ). H Outline is not confirmed for T Faintest with B ∅ and T Strongest , withno notable difference based on
OUT LINE . However, there aredifferences for T Faintest and with B Trees , for which P Fill , P Stripes × D Low , P Stripes × D High and P Dots × D Low have smaller opacityvalues (around 5% smaller) with O On than with O ∅ , i.e. the opacityranges are extended by around 5% (except for P Dots × D High ). H Background is not confirmed as there is no strong differencebetween any of the conditions based on
BKG . The results reveal additional aspects beyond hypotheses checking.First, different conditions lead to different opacity ranges. Opacityranges for P Fill are consistently slightly larger than for P Stripes × D Low , which are larger than P Dots × D Low , which are larger than P Stripes × D High , which are larger than P Dots × D High .Second, the CIs for T Strongest are wider than for T Faintest , i.e. thereis higher variance and uncertainty with T Strongest than with T Faintest .Third, Figure 4 shows the deviation from participant averagecompletion time for both tasks (instead of absolute completion timesthat are meaningless with the remote setup). A configuration ledto faster completion if its CI falls on the left side of the 0 baseline,and to slower completion if it falls on the right side. We notice that
BKG strongly affects participant completion time, with nearly allconditions with B Trees resulting in longer completion times thanaverage. We also observe that P Fill and P Stripes × D Low might resultin slightly shorter times than P Dots × D High . P Fill as their favorite pattern. P Fill alsoreceived the strongest scores for performing both T Faintest (12 partic-ipants strongly positive) and T Strongest (13 strongly positive), and forconfidence in their answers (14 strongly positive). P Stripes receivedmitigated scores for T Faintest (7 positive, 6 negative) but was betterranked for T Strongest (10 positive, 3 negative), and made participantsquite confident in their answers (12 positive, 2 negative). P Dots wasthe least preferred pattern, with 7 positive and 6 negative for T Faintest ,7 positive and 6 negative for T Strongest , and 6 positive and 5 negativefor confidence in their answers. Participants found that O On helpedfor completing both T Faintest (13 positive) and T Strongest (11 positive),and increased their confidence in their answers (13 positive). D ISCUSSION
We did not expect that increasing the density of a pattern would alsoincrease the required opacity value, with the effect being stronger for T Faintest . It is possible that a more complex pattern makes it morechallenging to mentally reconstruct the information about two layers(the overlay and the 3D scene underneath) from a single image, thusrequires stronger visual stimuli and salience, i.e. higher overlayopacity value. Completion times conform with this observation.The time required to set the opacity for D High is in almost all caseslonger than for D Low , which may be due to longer times spent tomentally reconstruct the overlay. Since for T Strongest the pattern ismore visible the effect is not as strong as for T Faintest . Some overlay configurations have wider opacity ranges than others. P Dots has relatively smaller, shifted upwards ranges, compared to P Fill and P Stripes (i.e. higher minimum and maximum bounds). Wethink this is again due to the complexity of the pattern, and thathigher opacity values for the overlays with P Dots make the distinctionof the two layers (overlay and 3D scene underneath) easier.Figure 5 shows differences in distributions of opacity values usedfor each
PAT T ERN . All three plots have two inflexion points, fortheir minimum and maximum bound of opacity range, but theiroverall shapes differ. The extent of these opacity ranges, combinedwith their degree of bimodality (how separable the minimum boundis from the maximum bound) can guide the selection of renderingparameters according to the data to visualize. For example, an over-lay configuration with a small extent and a low degree of bimodalitywould be appropriate to visualize a binary data dimension, e.g enableor disable the overlay; a high degree of bimodality is appropriate fora quantitative data dimension; and a large extent is appropriate for aquantitative data dimension that requires a high resolution.
Our results show that an outline can help increase the acceptableopacity range of an overlay, especially on the lower opacity bound.With an outline the pattern is no longer solely responsible for estab-lishing the structure of the overlay, it serves as an additional internalsupport and can be less salient. Outlines are particularly useful whenthere are objects in the scene, increasing the range of acceptableopacity values (lower bound) by around 5% in that case. Outlinesare less useful for T Strongest , because the region of interest is alreadysufficiently visible due to the salience of the pattern itself.We found that participants were slower with objects (trees) inthe background, for all conditions we tested. We explain this resultbecause the rendering of an overlay is partially occluded by theobjects inside. It results in a more complex overlay shape withuncertain boundaries that takes more time to reconstruct mentally.Several participants found that P Stripes helped identify the struc-ture of the terrain. However, we noticed that this happened mostlywhen the stripes were parallel to the slope of the terrain. One way toimprove structural visibility would be to use grid-like patterns thatwould assist in understanding the gradient in two dimensions.
Our empirical results provide a starting point to establish how to de-sign overlaid visualizations of abstract data in realistically rendered3D environments. This first step necessarily comes with limitationsin terms of generalizability of the results.First, the study measures people’s preferences for opacity rangesinstead of visual performance. This is a subjective measure thatallows us to derive the most comfortable opacity ranges for the user.In the future we plan to conduct quantitative perceptual studies thatwill help us measure more objectively visual performance.
OPACITY
Fully transparent Fully opaque P Fill P Dots P Stripes
Figure 5: Distributions of all opacity values for each
PAT T ERN . Second, we used a landscape with and without trees, but otheraspects of the terrain need to be investigated: what if the texture ofthe terrain is light, like snow, or dark, like grass or forest? What ifit is uniform, like an ocean, or has many topological features, likea mountain? There are also many more possible design variationsof the overlays to study. These include the color of the overlay, thecontrast between the dominant color of the terrain and the color ofthe overlay, the orientation of the patterns, the distribution of pointsand lines in the patterns and the point shape and radius.The world of computer graphics offers a wide array of additionalrendering parameters that might not be traditionally used whenvisualizing abstract data on 2D charts (or even avoided by fear ofcreating chart junk), but that could be relevant for overlays in 3Denvironments. This includes the impact of color, contrast, shadingor reflections. These additional considerations may be valuable tomake it easier to discriminate the overlay from the background whenthey are of similar color. In our study we kept the rendering assimple as possible to avoid artifacts or distractions, but one couldfor example look at casting shadows and incorporating reflections.Another important aspect to study is that of intersecting overlays.Having multiple (overlapping) RoI to analyze simultaneously iscommon within industrial applications. How to display multipledata sets is a current challenge that also calls for additional research.
ONCLUSION
We have provided the first set of empirically determined acceptable(subjective) opacity ranges to use when adding overlays on top ofa photorealistic 3D environment. Through our study we observedparticipants set the lower and upper limits for opacity values forthree different patterns (filled, striped, and dotted patterns), twopattern densities, the presence or not of a solid outline, and twotypes of background (blank and with trees).We found that the range of acceptable opacity values is roughlybetween 20-70%, with ranges shifting up or down slightly accordingto the overlay configurations. While patterns like dots are mostlyconsidered as distracting, we found that striped patterns can helpbetter understand the curvature of the underlying terrain.How to coalesce data visualization with realistic and interactiverendering in a unified visual context is an open research questionwith practical industrial impacts. Rather than providing definiteanswers to this open question, our initial inquiry provides manydirections worth studying to better understand how to representoverlays on top of complex, realistic 3D scenes. We look forwardto continuing this exploration at the junction of information andscientific visualization, and realistic computer graphics rendering. A CKNOWLEDGMENTS
This work was funded by NSERC, the Carl Zeiss Foundation andthe Federal Ministry for Economic Affairs and Energy of Germany.
EFERENCES [1] Cesium Stories. https://cesium.com/cesium-stories/ , 2020.Accessed: 2020-04-28.[2] ESRI ArcGIS Mapping and Analytics System. , 2020. Accessed: 2020-04-28.[3] TeamViewer Remote Desktop Software. , 2020. Accessed: 2020-04-28.[4] American Psychological Association.
The Publication manual of theAmerican psychological association . Washington, DC, 6th ed., 2013.[5] L. Bartram and M. C. Stone. Whisper, don’t scream: Grids and trans-parency.
IEEE Transactions on Visualization and Computer Graphics ,17(10):1444–1458, 2011.[6] W. S. Cleveland and R. McGill. Graphical perception: Theory, ex-perimentation, and application to the development of graphical meth-ods.
Journal of the American statistical association , 79(387):531–554,1984.[7] C. Collins, S. Carpendale, and G. Penn. Visualization of uncertainty inlattices to support decision-making. In
EuroVis , pp. 51–58, 2007.[8] G. Cumming and S. Finch. Inference by eye: Confidence intervals andhow to read pictures of data.
American Psychologist , 60(2):170, 2005.[9] M. Dodge, M. Mary, and M. Turner.
Geographic Visualization: Con-cepts, Tools and Applications . John Wiley & Sons, 2011.[10] F. H. Durgin and D. R. Proffit. Visual learning in the perception oftexture: Simple and contingent aftereffects of texture density.
Spatialvision , 9(4):423–474, 1996.[11] M. D’Zmura, P. Colantoni, K. Knoblauch, and B. Laget. Color trans-parency.
Perception , 26(4):471–492, 1997.[12] A. Gooch, B. Gooch, P. Shirley, and E. Cohen. A non-photorealisticlighting model for automatic technical illustration. In
Proceedings ofthe 25th Annual Conference on Computer Graphics and InteractiveTechniques , SIGGRAPH ’98, p. 447452. Association for ComputingMachinery, New York, NY, USA, 1998. doi: 10.1145/280814.280950[13] H. Haase, M. Bock, E. Hergenr¨other, C. Kn¨opfle, H.-J. Koppert,F. Schr¨oder, A. Trembilski, and J. Weidenhausen. Meteorology meetscomputer graphicsa look at a wide range of weather visualisations fordiverse audiences.
Computers & Graphics , 24(3):391–397, 2000.[14] B. M. Hughes. Just noticeable differences in 2d and 3d bar charts: Apsychophysical analysis of chart readability.
Perceptual and motorskills , 92(2):495–503, 2001.[15] V. Interrante, H. Fuchs, and S. Pizer. Illustrating transparent surfaceswith curvature-directed strokes. In
Proceedings of Seventh AnnualIEEE Visualization’96 , pp. 211–218. IEEE, 1996.[16] M. Krzywinski and N. Altman. Points of significance: error bars.
Nature methods , 10(10):921–922, 2013.[17] K. Lawonn and B. Preim. Feature lines for illustrating medical surfacemodels: Mathematical background and survey. In
Visualization inMedicine and Life Sciences III , pp. 93–131. Springer, 2016.[18] K. Lawonn, N. N. Smit, K. B¨uhler, and B. Preim. A survey on mul-timodal medical data visualization. In
Computer Graphics Forum ,vol. 37, pp. 413–438. Wiley Online Library, 2018.[19] K. Lawonn, I. Viola, B. Preim, and T. Isenberg. A survey of surface-based illustrative rendering for visualization. In
Computer GraphicsForum , vol. 37, pp. 205–234. Wiley Online Library, 2018.[20] A. Mairena, M. Dechant, C. Gutwin, and A. Cockburn. A baselinestudy of emphasis effects in information visualization. In
Proceedingsof Graphics Interface 2020 , p. 13 pages. Toronto, Ontario, 2020.[21] D. Mart´ın, G. Arroyo, A. Rodr´ıguez, and T. Isenberg. A survey ofdigital stippling.
Computers & Graphics , 67:24–44, 2017.[22] J. E. McGrath. Methodology matters: Doing research in the behavioraland social sciences. In
Readings in Human–Computer Interaction , pp.152–169. Elsevier, 1995.[23] F. Metelli. The perception of transparency.
Scientific American ,230(4):90–99, 1974.[24] T. Munzner.
Visualization Analysis and Design . A K Peters & CRCPress, 2014.[25] B. Schleich, N. Anwer, L. Mathieu, and S. Wartzack. Shaping thedigital twin for design and production engineering.