Analyzing and Optimizing Video Quality in the new H.265 (HEVC) Standard

摘要

The emerging H.265 High Efficiency Video Coding (HEVC) standard for video compression is geared towards providing high quality/high resolution video at low to moderate bit rates; thus providing a significant increase in compression efficiency over existing standards like H.262/263/264. As is typically the case, increased compression results in decreased coding redundancy. Hence from a video quality perspective, of interest would be the nature of potential coding artifacts as well as the profile of video failures or dropouts arising due to uncorrected coding errors. In the latter case, the occurrence of these uncorrectable errors would depend upon external factors like the channel SNR or Eb/No profile, coupled with the source coding efficiency (or redundancy); the latter being governed by the opposing constraints of picture quality versus allowable bit rate or channel bandwidth. The initial portion of this work is focused on dropout profiles due to randomly occurring bitstream errors. The geometries of the manifested dropout regions are quite noticeable as compared to those observed in traditional slice based video syntax for the same content, under a similar random error profile. This uniqueness impacts the spatial dropout detection algorithms to be used. In the latter part of this work, video quality for a variety of content applications (like sports, news/talk shows, animation, etc...) and under varying encoder parameter profiles is monitored. The goal is to optimize the HEVC encoding parameters such as Transform Unit size, Coding Unit size and Maximum Partition Depth for video quality, for each chosen content application.