Quality Assessment of MPEG‐4 Compressed Videos

The recent MPEG‐4 Part 10/H.264 AVC standard enables improved compression rates without compromising on visual quality. ‘Visual quality’ here refers to the quality of a video as perceived by a human observer. It is this aspect of visual quality that we focus upon in this chapter. It is widely agreed that the most commonly used mean squared error (MSE) correlates poorly with the human perception of quality. MSE is a full reference (FR) video quality assessment algorithm (VQA). FR VQA algorithms are those that require both the original as well as the distorted videos in order to predict the perceived quality of the video. Recently proposed FR VQA algorithms have been shown to correlate well with human perception of quality. However, a practically implementable solution remains evasive. In this chapter, we detail one possible approach to real‐time quality assessment, developed specifically for MPEG‐4 compressed videos. This algorithm leverages the computational simplicity of the structural similarity index (SSIM) for image quality assessment (IQA), and incorporates motion information embedded in the compressed motion vectors from the H.264 compressed stream to evaluate visual quality. We detail the algorithm and demonstrate its performance on the popular Video Quality Experts Group (VQEG) FRTV Phase – I dataset. Further, we describe a subjective study that we have undertaken specifically for H.264 compressed videos. We compare the performance of leading VQA algorithms on this database and make the database available free‐of‐cost for researchers in order to further the field of VQA for MPEG‐4 compressed videos.