Jonscher M (2018)
Publication Language: English
Publication Type: Thesis
Publication year: 2018
Publisher: Dr. Hut
City/Town: München
ISBN: 978-3-8439-3839-6
Video data with both a high spatial and a high temporal resolution is desired for many applications. The processing throughput of state-of-the-art cameras, however, is limited to a certain number of pixels per second which leads for most of these cameras to a trade-off between acquiring a video sequence at a high spatial resolution but a low temporal resolution or vice versa. In order to overcome these limitations, a non-regular sampling using a subsequent reconstruction algorithm may be applied.
This thesis focuses on the reconstruction of non-regularly sampled image and video data. Various enhancements for the applied reconstruction algorithm are introduced that improve the reconstruction quality on the one hand or reduce the computational complexity on the other hand. Techniques that exploit temporal or spatial correlations when using a non-regular sampling sensor in multi-frame or multi-view scenarios show promising results. Additionally, a novel dynamic non-regular sampling strategy is presented that allows the readout of varying sampling patterns over time and in combination with the aforementioned reconstruction algorithm, significant gains in reconstruction quality may be achieved.
APA:
Jonscher, M. (2018). Frequency Selective Reconstruction of Non-Regularly Sampled Image and Video Data (Dissertation).
MLA:
Jonscher, Markus. Frequency Selective Reconstruction of Non-Regularly Sampled Image and Video Data. Dissertation, München: Dr. Hut, 2018.
BibTeX: Download