Abstract

Compression of digital video introduces artifacts (i.e., typical types of degradations), such as Blocking, Blurring, Ringing. Nowadays, digital video systems have been applied widely to various areas, and this has led to a rising demand for video quality metrics. In the past ten years, many automatic metrics of video quality have been developed; among them, the Wolf-Pinson metric is the better one. Based on the Wolf-Pinson metric, an Artifact-Measured-based Automatic Metric (AMAM) is proposed in this thesis, which measures four artifacts (Blocking, Ringing, Blurring and Block Flashing). Among them, Block Flashing is an artifact not previously measured, in the proposed metric, the Masking property of Human Visual System (HVS) is taken advantage to measure the Blocking and Ringing in order to improve performance. In this thesis, the performance of the automatic metrics is evaluated by using the Pearson linear correlation coefficient, Mean Square Error and Spearman rank-order correlation coefficient. The simulation results indicate that the performance of the AMAM is significantly better than those of other simulated metrics (including the Wolf-Pinson metric) for MPEG-2 test video sequences.