University of Thessaly (UTH)
Algorithms and advanced methods for video coding in large scale systems
Abstract
dc:descriptionThe new era of 4K and 8K (UHD) resolutions cameras and TV sets emerged from the ever increasing demands for higher video resolution. Video is by far the "biggest" Big Data, stretching network and storage capacity to their limits. In order to handle this situation, video compression has been an active field of study for many years, producing output of huge commercial interest. One of the most successful and widely used video coding standards is H.264/AVC. Even if H.264/AVC was a wide marketing success, its weaknesses and limitations in tacking UHD resolutions were made apparent relatively early. For this reason, new video coding standards were launched as successors to H.264/AVC. The most eminent examples of the new generation video coding standards are the High Efficiency Video Coding (HEVC), the recently launched AV1 and the currently developing Versatile Video Coding (VVC). Each one of the video coding standards was designed with the primary target of outperforming the prior state of the art standards. Therefore, each newcomer introduced new coding tools that increases computational complexity both at the encoder and at the decoder side, and trade-off mechanisms between complexity and compression efficiency were presented at the design steps of every standard. Both the high resolution demands and the increased computational complexity of video coding standards, result to time consuming and resource demanding tasks. Harnessing the high computational demands is done by exploiting parallelization at various levels. In this thesis, we tackle both of the aforementioned aspects, namely, harnessing computational overhead through parallelism and exploiting the potential of resource consumption reduction at a high scale. Concerning the first aspect (parallelism), we focus on a block level coarse grained parallel approaches, which split a frame into groups of blocks and consider each group as an independent task. In particular, tile partitioning is considered whereby a frame is split into horizontal and vertical zones, the intersections of which form tiles. We propose and evaluate algorithms that define tile partitions under various scenarios with the aim of balancing the resulting load in the available processors. Concerning resource consumption reduction at a high scale, the improved compression ratio achieved by HEVC, compared to H.264/AVC, is the key enabling factor. Furthermore, we tackle issues related to H.264/AVC-HEVC transcoding. Towards this aspect, two paths are explored. In the first path, we propose a method for estimating regions of high complexity within a frame, based on information that can be extracted from H.264/AVC headers. Such information is important in the case of a parallel transcoder, as it can be used for efficient load balancing. In the second path, we tackle the problem of scheduling transcoding processes at a massive scale so as to maximize resource benefits while minimizing operational costs.
Degree
thesis:*- Grantor dc:publisher
- University of Thessaly (UTH)
- Year dc:date
- 2020
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Papadopoulos, Panagiotis
- Παπαδόπουλος, Παναγιώτης
Subjects
dc:subject × 17- Συμπίεση βίντεο
- Κωδικοποίηση βίντεο
- Παραλληλοποίηση
- Διακωδικοποίηση βίντεο
- Yπολογιστικό νέφος
- Video coding
- HEVC
- Tiles
- Slices
- Transcoding
- Video compression
- Cloud computing
- Parallel computing
- Φυσικές Επιστήμες
- Επιστήμη Ηλεκτρονικών Υπολογιστών και Πληροφορική
- Natural Sciences
- Computer and Information Sciences
Rights
- Language dc:language
- eng
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/47597