Inter-Sequence Error Concealment Techniques for Multi-Broadcast TV Reception

Mobile reception of digital TV signals often suffers from lost slices due to suboptimal channel characteristics. In this contribution we propose novel inter-sequence error concealment techniques which reconstruct lost blocks of distorted high-resolution TV signals by inserting error-free blocks of low-resolution reference signals. The proposed methods are well-suited for future automotive multi-broadcast receivers. As the reference signals are typically delayed, scaled and possibly cropped versions of the distorted TV signals, spatio-temporal image alignment is a crucial point. We first focus on pixel-based alignment utilizing a numerical optimization scheme. To reduce complexity a feature-based approach is introduced alternatively taking advantage of the scale-invariant feature transform. Simulations show that the proposed techniques can achieve supreme reconstruction quality outperforming conventional techniques by up to 15.0 dB
${\rm PSNR}_{\rm Y}$
. The pixel-based approach mostly reaches the maximum reconstruction quality for both raw and compressed video signals. Feature-based concealment performs slightly worse but reduces the computation time by a factor of up to 36.6 at a maximum loss of 1.0 dB. A novel hybrid concealment technique combining the proposed methods with a temporal approach further increases the reconstruction quality. Applying one simple and one sophisticated mode selection scheme, a gain of 3.2 dB and 4.6 dB can be achieved, respectively.