Hybrid NR Video Quality Metric with Decodable Payload

Abstract

In this paper, we propose a hybrid NR video quality metric with decodable payload. Although NR video quality metric has a strong need from the industry, developing these metrics has been a challenging task. Recently, hybrid models that use both processed video sequences and bit-stream data have shown promising results. In particular, from bitstream data, useful information can be extracted for video quality estimation, which includes codec information and impairment due to transmission errors. In the proposed method, we used packet information from the bit-stream data and estimated transmission impairment by decoding the payload to estimate the video quality. The proposed method was tested for the H.264 standard and it showed promising results. Also, the proposed method was independently evaluated using large datasets and it produced good performance.