Modeling of Perceptual Video Quality Considering Impact of Spatial, Temporal, and Amplitude Resolutions

In networked video applications where the sustainable bandwidths vary greatly among the receivers and in time, one must be able to code a video or adapt a precoded-bitstream to a wide range of bit rates. It is critical to choose appropriate frame size (FS), frame rate (FR) and quantization stepsize (QS) to optimize the perceptual quality for a target bit rate. Furthermore, it is important to understand the variation effect of these parameters, so that appropriate constraints can be imposed when adapting these parameters in response to bandwidth changes. However, well-established models that relate the perceptual quality with the spatial, temporal, and amplitude resolutions (STAR) and the variation of STAR do not exist today.;In this dissertation, we conducted three main subjective quality experiments. First one examines the impact of FR and QS on perceptual quality of a video for laptop devices. We propose to use the product of a spatial quality factor that assesses the quality of decoded frames without considering the frame rate effect and a temporal correction factor, which reduces the quality assigned by the first factor according to the actual frame rate. We find that the temporal correction factor follows closely an inverted falling exponential function of FR, whereas the quantization effect on the coded frames can be captured accurately by a sigmoid function of the PSNR or by an exponential function of QS. The complete model correlates well with the subjective ratings with a Pearson Correlation Coefficient (PCC) of 0.98 when parameters are obtained by least square fitting with the subjective ratings and a PCC of 0.97 when model parameters are predicted from the content features.;The second experiment investigates the impact of STAR on the perceptual quality of a compressed video. Subjective quality tests were carried out on the TI Zoom2 mobile development platform (MDP). Subjective data reveals that the impact of SR, TR and QS can each be captured by a function with a single content-dependent parameter. The joint impact of SR, TR and QS can be modeled by the product of these three functions with only three parameters. The complete model correlates very well with the subjective ratings with a PCC of 0.99 when parameters are obtained by least square fitting and a PCC of 0.98 when model parameters are predicted from the content features. We further validate our model on several datasets reported from other works and the accuracy of our model (in part or in whole) on these datasets is still promising.;The third experiment explores the impact of periodic frame rate and QS variation on perceptual video quality. Among many dimensions of FR/QS variation, as a first step we focus on videos in which two FR's, or QS's, alternate over fixed intervals. According to the observation and data analysis of the test results, we propose models that characterize the quality degradation with respect to FR, QS and bit rate variation. These quality models can help to make appropriate decisions for encoder adaptation when transmitting video over networks with fluctuating bandwidth.