An investigation on perception of computer generated surface materials and the influence of environment complexity.

En Undersøgelse af Opfattelsen af Computer Genereret Overfalde Materialer og Indflydelsen af Scene Kompleksitet.

Lumbye, Anders

4. term

Medialogy, Master

2014

2014-05-28

67

Computergenererede billeder (CGI) er i dag så realistiske, at de kan ligne fotografier. I arkitektoniske visualiseringer forsøger kunstnere at genskabe hverdagens små uperfektheder – fingeraftryk, ridser og snavs – for at undgå et klinisk udtryk og nå en hyperrealistisk stil. Men så realistiske billeder kan også være vildledende for kunder, hvis det færdige resultat ser anderledes ud end forventet. En medvirkende årsag kan være, at overfladematerialer i CGI ikke er tilstrækkeligt validerede. Dette projekt udviklede en ny tilgang til at validere, hvordan virkelige overfladematerialer gengives i CGI, afprøvet i to miljøer med forskellig kompleksitet (lavt vs. højt semantisk indhold), med fokus på menneskelig perception. Formålet var at undersøge, om folk bedømmer lighed mellem materialer forskelligt i en simpel testscene sammenlignet med en mere kompleks, kontekst-rig scene, og om tærsklen for at acceptere noget som “tilstrækkeligt lignende” er den samme i begge miljøer. Tre materialetyper indgik: et stærkt spejlende (meget blankt) materiale (whiteboard), et diffust (mat) materiale (Post-it-note) og et glansfuldt (halvblankt) materiale (bord). Hver blev kalibreret så tæt som muligt på en virkelig reference, og der blev lavet små, kontrollerede afvigelser for at skabe serier af prøver. Bedømmere kunne skelne forskellene i begge miljøer. I det lavt komplekse miljø var der større enighed mellem bedømmerne, og grupperingerne for det spejlende og det diffuse materiale var tydeligere end i det komplekse miljø. Det glansfulde materiale gav derimod få klare grupperinger i begge miljøer, hvilket tyder på en bredere accept af variation for denne type overflade. Resultaterne peger på, at scenekompleksitet påvirker, hvor konsekvent mennesker vurderer materialers realisme i CGI, og understreger behovet for perceptuel validering for at undgå misforståelser med kunder.

Computer-generated imagery (CGI) is now so realistic that it can be hard to distinguish from photography. In architectural visualization, artists add small imperfections—fingerprints, scratches, dirt—to avoid a clinical look and achieve a hyper-realistic style. However, such realism can also mislead clients if the finished result differs from what the images suggested. One reason may be insufficient validation of surface materials in CGI. This project developed a new approach to validate how real surface materials are represented in CGI, tested in two environments with different complexity (low vs. high semantic content), with human perception at the center. The goal was to see whether people judge material similarity differently in a simple test scene compared with a more complex, context-rich scene, and whether the threshold for accepting something as “similar enough” is consistent across both. Three material types were examined: a highly specular (very shiny) material (whiteboard), a diffuse (matte) material (Post-it note), and a glossy (semi-shiny) material (table). Each was calibrated to match a real reference as closely as possible, then small, controlled deviations were introduced to create sets of slightly different samples. Assessors could discern differences in both environments. In the low-complexity scene, assessors agreed more with each other, and their groupings for the specular and diffuse materials were clearer than in the complex scene. The glossy material produced few clear groupings in either environment, suggesting a wider tolerance for variation for this type of surface. These results indicate that scene complexity influences how consistently people judge material realism in CGI and highlight the need for perceptual validation to avoid client misinterpretation.

[This abstract was generated with the help of AI]

Computer ; Generede ; billeder ; opfattelse ; overflade materialer ; BRDF ; kontekst

Download
View record in AAU Student Projects