A Correlation-Based No-Reference Packet-Loss Metric

In the last few years, there has been a growing interest in automatic methods (metrics) that blindly estimate the quality of digital videos, specially in applications where the videos are digitally transmitted over wired or wireless channels and the original (reference) video signal is not available at the receiver. Currently, there are few methods that are able to identify and estimate the strength of temporal artifacts (e.g. packet-loss and jerkiness) introduced during a digital transmission. In this paper, we present a blind (no-reference) video quality assessment metric that estimates the impact of packet-loss artifacts on quality. The proposed algorithm performs a comparison of inter and intrablock correlations, considering blocks of sizes 88, 1616, and 3232. The proposed metric was tested on a database containing videos with packet-loss artifacts with different strengths and time durations. Results show that the proposed metric is able to estimate the impact that packet-loss artifacts have on the overall perceived quality, outperforming several metrics available in the literature.