Pixel domain referenceless visual degradation detection and error concealment for mobile video

Abstract

In mobile video applications, where unreliable networks are commonplace, corrupted video packets can have a profound impact on the quality of the user experience. In this paper, we show that, in a wide range of operating conditions, selectively reusing data resulting from decodable errorneous packets leads to better results than frame copy. This selection is guided by a novel concept that combines motion estimation and a measure of blocking artifacts at block edges to predict visual degradation caused by the decoding of erroneous packets. Simulation results show that, by using the proposed solution, the H.264/AVC JM reference software decoder can select the best option between frame copy and the erroneous frame decoding in 82% of test cases. We also obtain an average gain of 1.95 dB for concealed frames (when they differ from those concealed by the JM decoder).

Video Error Correction Using Soft-Output and Hard-Output Maximum Likelihood Decoding Applied to an H.264 Baseline Profile

“…However, we believe that sending those packets to the video decoder, along with additional information such as soft-output information, is a better strategy. Although the number of errors is unknown, corrupted packets often contain information that can be exploited as shown in [18], [19], [20], [21], [22]. Fig.…”

“…where each term in the product, as well as the final term (the one outside the product) is evaluated using (22). However, the address at which the conditional probability will be evaluated is specified by the discrete random variable ADDR.…”

“…The idea was to retrieve intact information from the corrupted packets prior to applying error concealment in order to reconstruct higher quality images. More recently, Superiori [20], Farrugia [21], and Trudeau [22] have demonstrated that similar error detection approaches also work for the H.264 standard.…”

Performance evaluation of maximum likelihood decoding combined with error resilient video coding

“…In practice, network congestion results in packet loss while attenuation, fading, and the likes result in corrupted packets. These partially damaged packets may however contain valuable information, which is useful for improving the visual quality of the reconstructed video [21][22][23]. The goal of video error correction is to use these corrupted packets to recover the originally sent values rather than assuming that the whole packet is lost.…”

A Maximum Likelihood Approach to Video Error Correction Applied to H.264 Decoding

“…In fact, video communication systems belong to a class of applications that would benefit from using corrupted packets (Larzon et al, 1999;Trudeau et al, 2011). Accessing those packets…”