PhD Alumni

Naccari Matteo

Present position: Senior Research Engineer BBC R&D

Thesis title:  Innovative video coding tools for error resilience, automatic quality monitoring and transcoding
Advisor:  Stefano Tubaro
Research area:  Digital Signal Processing
Thesis abstract:  
The thesis addresses the transmission of coded video contents by three different points of view: first, how to perform optimal transcoding in order to allow the use of different coding formats for storing and transmitting video contents. Second, how to add redundancy to a coded content in order to increase its resilience to transmission errors and finally, how to automatically estimate the video quality experienced by the final user when transmission is subjected to errors. These points are addressed by means of the following technologies: video transcoding, robust coding and automatic quality monitoring. Video transcoding consists in the conversion of the bitstream X into another one Y by means of a system called transcoder. The thesis addresses the conversion from the H.264/AVC to MPEG-2 in order to allow the distribution of video contents to terminals MPEG-2 compliant (like standard Set-Top-Boxes). The second addressed point can be seen as a special case of transcoding: in fact, the coded bitstream can be modified in order to increase its robustness against channel errors. Moreover, the thesis addresses also the special case where a protection stream is added “on-top” of the original coded content by means of tools provided by the Distributed Source Coding (DSC) theory. Automatic quality monitoring, provides techniques and methodologies to estimate the video quality of coded and transmitted bitstreams. The estimation of the video quality can be performed either at the transmitter or the receiver side. The thesis addresses the estimate of the video quality degradation induced by channel errors at the receiver side. At this side, the estimation is complicated by the fact that the original video content is not available for comparison (no-reference approach). The thesis proposes a no-reference quality monitoring algorithm that explicitly models the distortion induced by channel errors in H.264/AVC compliant bitstreams. In addition to this, we also consider the case in which some small knowledge on the original video content is available at the receiver side (reduced-reference approach).