Datenbestand vom 15. November 2024
Tel: 0175 / 9263392 Mo - Fr, 9 - 12 Uhr
Impressum Fax: 089 / 66060799
aktualisiert am 15. November 2024
978-3-8439-3804-4, Reihe Ingenieurwissenschaften
Andrea Eichenseer Fisheye Video Processing Using Distortion-Corrected Block-Based Motion Compensation
226 Seiten, Dissertation Universität Erlangen-Nürnberg (2018), Softcover, A5
Surveying wide areas with a single camera is a typical scenario in surveillance and automotive applications. Conventional perspective camera lenses, however, are not able to cover very large fields of view. This is where ultra wide-angle fisheye cameras come into play. By employing fisheye lenses, the standard pinhole model is abandoned in favor of a fisheye projection that is able to accommodate fields of view that may well surpass 180 degrees.
The captured fisheye video sequences exhibit image characteristics that deviate notably from conventional rectilinear imagery as obtained by perspective pinhole cameras. These characteristics are not taken into account by traditional video processing techniques such as motion estimation and its subsequent compensation. This thesis therefore investigates fisheye video sequences in the context of motion estimation and various video processing tasks based thereon.
After introducing an extensive fisheye data set, a first analysis examines the performance of a state-of-the-art video compression scheme when faced with fisheye input sequences. The main part of the thesis is then dedicated to motion estimation based on block matching and the development of an adapted approach for fisheye video sequences. By incorporating suitable projections, the translational motion model that would otherwise only hold for perspective video sequences may be exploited, thus facilitating the motion search and yielding improved motion compensation results. Exhaustive simulations and experiments on synthetic and real-world fisheye video data demonstrate the merit of the adaptation.
Building upon fisheye motion estimation, four motion compensated video processing applications are investigated and adapted to provide improved results for fisheye input data. The considered applications include temporal error concealment, intermediate view synthesis, and temporal as well as spatial resolution enhancement.