Sparse recovery-based error concealment for multiview images

• Published in: 2015 International Workshop on Computational Intelligence for Multimedia Understanding (IWCIM) pp. 1 - 5
• Main Authors: Akbari, Ali, Trocan, Maria
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.10.2015
• Subjects: Correlation; Decoding; Error concealment; Image reconstruction; multiview images; Packet loss; sparse recovery; Wavelet transforms
• DOI: 10.1109/IWCIM.2015.7347067

Error concealment is a receiver-based post-processing technique that deals with the errors occur in a decoded signal during transmission in the packet loss environments, based on data available only at the decoder. In this paper, we propose a novel error concealment algorithm for the multiview images based on sparse reconstruction. Using a simple combination of the wavelet transform prior to transmitting the multiview data, the channel is considered like a random Gaussian sampling system so that the sparse recovery can be sufficiently used for the error concealment at the receiver side. Moreover, an efficient procedure for increasing the reconstructed quality based on an optical flow prediction is further proposed. Experimental results show a large improvement over the interpolation-based techniques in terms of PSNR in high frequency packet loss rates.