Summary:

The transmission of video content now accounts for the bulk of the world's data flows. The quality of this content is constantly improving, thanks to the deployment of networks capable of supporting more traffic at higher speeds, coupled with strategies aimed at reducing the information required to transmit a video sequence, through increasingly effective video compression techniques. Nevertheless, the visual quality of video content can be degraded for the end user when errors occur during transmission. A packet may be lost or corrupted during transmission, due to disturbances inherent in the broadcast channel. In order to recover the missing information, the corrupted packet can be retransmitted. However, this option is not always valid under real-time constraints, such as when broadcasting live video content, or in order not to increase the network load. It is then possible to implement error correction methods at the receiver to recover erroneous data. In this thesis, we propose receiver-level error correction methods that exploit Cyclic Redundancy Check (CRC) error detection codes for error correction. The methods we propose use the corrupted packet syndrome to compile an exhaustive list of error patterns that could have produced this syndrome, for a number of errors less than or equal to that defined as an input parameter. We propose several approaches to achieve this result, firstly using an arithmetic approach, based on logical operations performed on the fly and therefore not requiring memory storage. The second approach proposes an optimised table, in which the repetitive calculations of the arithmetic approach are stored efficiently and appropriately for the proposed method, enabling the correction to be carried out much more quickly, at the cost of memory storage. The validation of the correction is carried out by a two-stage process aimed at cross-referencing the list of candidates obtained with another error detection code, the checksum. The final stage involves checking the syntax of the encoded video stream by testing its decodability. Our methods have been tested on simulations of the transmission of video content compressed to H.264 and HEVC standards over 802.11p Wi-Fi and Bluetooth Low Energy channels. The latter case offers the most significant correction rates, resulting in the video being reconstructed virtually intact even when the quality of the transmission channel begins to deteriorate.

Abstract:

The transmission of video content now constitutes the bulk of the world's data flows. The quality of this content is constantly increasing, due to the deployment of networks that can support more traffic at higher data rates, coupled with strategies to reduce the information required to transmit a video sequence, through increasingly efficient video compression techniques. Nevertheless, the visual quality of video content can be degraded for the end user when errors occur during transmission. Indeed, a packet can be lost or corrupted during transmission, due to the inherent disturbances of the broadcast channel. In order to recover the missing information, a retransmission of the corrupted packet can be considered. However, this option is not always valid under real time constraints, such as when broadcasting live video content, or in order not to increase the network load. It is then possible to implement error correction methods at the receiver to recover erroneous data. In this thesis, we propose error correction methods located at the receiver, exploiting Cyclic Redundancy Check (CRC) error detection codes for error correction. Our proposed methods use the corrupted packet syndrome to compile an exhaustive list of error patterns that could have produced this syndrome, for a number of errors less than or equal to the one defined as an input parameter. We propose several approaches to achieve this result, first using an arithmetic approach, based on logical operations performed on the fly and thus not requiring memory storage. The second approach proposes an optimized table, in which the repetitive computations of the arithmetic approach are stored in an efficient way appropriate to the proposed method, allowing a much faster execution of the correction, at the cost of memory storage. The validation of the correction is done by a two-step process, aiming at crossing the list of the obtained candidates with another error detection code, the checksum. The final step aims at checking the syntax of the encoded video stream by testing its decodability. Our methods have been tested on simulations of transmission of video content compressed according to H.264 and HEVC standards over 802.11p Wi-Fi and Bluetooth Low Energy channels. The latter case offers the most significant correction rates, resulting in the video being reconstructed almost intact even when the quality of the transmission channel starts to deteriorate.