Video Transmission Artifacts Detection Using No-Reference Approach

Abstract

In real-time (RT) applications that include transmission of digital video, different artifacts (caused by compression and transmission processes) can appear in video received at the end-user side. In order to ensure high level of end-user Quality of Experience (QoE), video application/service providers have to continuously measure and monitor the quality of perceived video. Since in RT video applications uncompressed video is unavailable at the receiver side, artifacts detection as well as video quality assessment (VQA) are often performed using no-reference (NR) approach. In this paper we present a novel NR algorithm that efficiently detects packet loss (PL) artifacts in received video frames, called Packet Loss Detection Algorithm (PLDA). The proposed PLDA operates only on pixel values of the processed video frame and it requires no additional information about processed video. The performance of the proposed PLDA is compared to that of other existing PL detection algorithm on video sequences of significantly different content, in which distinct error concealment methods are used to conceal errors caused by PL. The results show that PLDA outperforms other tested algorithm when detecting PL artifacts in network transmitted video and that it is very robust in terms of different content types and error concealment methods. Additionally, proposed PLDA is capable of processing up to 25 frames per second (FPS) of Full HD video in RT and thus it is suitable for usage in RT video transmitting applications.