AAU Student Projects - visit Aalborg University's student projects portal
A master's thesis from Aalborg University
Book cover


A Study in Perceived Believability: Utilizing Visual Movement to alter the Level of Detail

Author

Term

4. term

Education

Publication year

2013

Submitted on

Pages

67

Abstract

This thesis investigates whether visual motion can steer viewers’ attention and thereby allow a reduction of detail in other CGI elements without diminishing perceived believability. The rationale is that the human visual system has limits and that attention can be both goal‑driven and stimulus‑driven, with motion being a strong cue that draws focus. Based on this, the core question is whether detail can be lowered in parts of a scene when another object captures attention, so the change goes unnoticed and the scene remains believable while reducing render time. The thesis outlines an approach that combines live action and CGI, manipulating motion and level of detail and collecting viewer ratings of believability. It describes a production pipeline including modeling, texturing, HDRI lighting, matchmoving, rendering, compositing, and the use of sound and a simple narrative to anchor elements in the scene. The experimental planning references factorial design, a Latin square, and ANOVA to analyze participant responses. Specific findings are not included in this excerpt; the focus here is to motivate the problem, define key concepts, and set out methods and success criteria to assess whether detail can be reduced without harming perceived believability.

Denne afhandling undersøger, om visuel bevægelse kan styre seerens opmærksomhed og dermed muliggøre en reduktion af detaljeniveauet i andre CGI‑elementer uden at svække den oplevede troværdighed. Baggrunden er, at den menneskelige visuelle system har begrænsninger, og at opmærksomhed både kan være målstyret og stimulusdrevet, hvor især bevægelse kan trække fokus. Med udgangspunkt i denne viden formuleres et problem: Kan man sænke detaljegrad i dele af en scene, når et andet objekt tiltrækker opmærksomheden, så det ikke bemærkes og scenens troværdighed bevares, samtidig med at render‑tid reduceres? Afhandlingen skitserer en tilgang, hvor live action og CGI kombineres, og hvor bevægelse og detaljeniveau manipuleres for at måle seerens vurdering af troværdighed. Der beskrives en pipeline med modellering, teksturering, HDRI‑lys, matchmoving, rendering og komposition samt brug af lyd og en enkel fortælling for at forankre elementerne i scenen. Den eksperimentelle planlægning nævner faktorforsøg, latinsk kvadrat og ANOVA til at analysere deltagernes ratings. Resultater præsenteres ikke i det udvalgte uddrag; fokus her er at motivere, definere begreber og opstille metode og succeskriterier for at vurdere, om detaljer kan reduceres uden at påvirke den oplevede troværdighed.

[This apstract has been generated with the help of AI directly from the project full text]