Masterarbeit MSTR-2025-108

Bibliograph.
Daten
Wurm, Mika: Object warping and view synthesis in a layered deep framebuffer.
Universität Stuttgart, Fakultät Informatik, Elektrotechnik und Informationstechnik, Masterarbeit Nr. 108 (2025).
47 Seiten, englisch.
Kurzfassung

This work presents a rendering method for extrapolating multiple frames of keyframe animated objects. It uses the Mojo framework as a basis which allows for multi-frame extrapolation of static scenes with a moving camera. In this work, an extension is developed which enhances the extrapolation capabilities of the framework to animated objects. To that end, a modified version of the layered framebuffer from the Mojo framework is devised, which supports the individual processing of the animated objects. In addition, a new reprojection method is developed. It combines the changing model matrix of an animated object with the camera movement during the extrapolated frames to create novel views of the animated object. The resulting framework achieves the overall goal of fluid object movement in the extrapolated frames. The analyzed image metrics of the test scenes show similar values to the results from the base Mojo framework. Translations of objects yield accurate warping while rotations are more prone to visible image stretching. A lot of the slight inaccuracies are not visible in motion. The additional performance overhead increases with more animated objects. The tests show frame time increases of to 44%. A necessary framework setting for achieving reasonable animated warping quality significantly limits the effective speed to a recorded maximum of 56 FPS with a single animated object. Additional work should relieve this performance restriction.

Volltext und
andere Links
Volltext
Abteilung(en)Universität Stuttgart, Institut für Visualisierung und Interaktive Systeme, Visualisierung und Interaktive Systeme
BetreuerSchmalstieg, Prof. Dieter; Künzel, Sebastian
Eingabedatum18. Mai 2026
Neuer Report   Neuer Artikel   Neues Sammelwerk   Institut   Informatik