Hybrid No-Reference Video Quality Models for H.264 with Encrypted Payload

Abstract

In this paper, we propose a hybrid no-reference video quality measurement method for H.264. Although there is a strong need from the industry for no-reference models that don’t require reference signals, developing reliable no-reference models for perceptual video quality assessment has been a challenging goal. On the other hand, in most digital video transmission, encoded video signals are transmitted in video packets, which can provide valuable information about coding parameters and transmission errors. The proposed hybrid noreference models uses both the information available from video packets and decoded signals to estimate coding artefacts and impairments caused by transmission errors. Sometime, the payloads of video packets may be encrypted to protect contents. In such case, the available information is rather limited. The proposed method was independently tested against large databases and shows good performance.