AN OBJECTIVE EVALUATION OF CODECS AND POST-PROCESSING TOOLS FOR 8K VIDEO COMPRESSION

Abstract

With the deployment of the latest Ultra High Definition Television (UHDTV) system, Quality of Experience (QoE) of users is expected to be improved through the introduction of new features to the existing High Definition Television (HDTV) system, such as High Dynamic Range (HDR), wider color gamut, High Frame-Rate (HFR) and higher spatial resolutions including 4K (3840x2160) and 8K (7680x4320).

The delivery of such video formats on current broadcast infrastructures is a real challenge and requires efficient compression methods to reach the available bandwidth while ensuring high video quality.

On the other hand, with the outstanding performance of AI-based spatial up-scalers, TV manufacturers have recently included them as a feature in their products. As 8K contents are still uncommon, it can be useful to reconstruct high-quality 8K from a lower resolution to exploit the screen capabilities. Moreover, these filters can also be used to reduce the high bitrate requirements of 8K by downscaling the original 8K signal before encoding and reconstruct it after decoding.

In this paper, we propose to evaluate different video coding scenarios for 8K compression using VVC and HEVC standards through an objective study. Tested configurations include: 8K source encoded with HEVC, and 8K and 4K sources encoded with VVC and then upscaled with two methods: a Lanczos filter and a deep-learning-based Super-Resolution method called SRFBN. All configurations are tested on a set of 8K sequences using the verification model of the VVC and HEVC standards.

Introduction

With the advent of new immersive video formats, High Dynamic Range (HDR), HFR (High-Frame Rate), and UHD (Ultra-High Definition) are among the main improvements that are currently and progressively introduced in recent deployments of television networks. The ultimate phase of the migration to the new video formats, a.k.a. UHD-2 [1], considers the introduction of the 8K definition (7680x4320), in addition to existing formats like 4K (3840x2160) or HD (1920x1080). By increasing the number of pixels, a finer signal digitization is possible, bringing more details into the image, and thus increasing the Quality of Experience (QoE) for the users. However, high bandwidth capacities are required to transmit 8K estimated to 80Mbps with HEVC encoder [2]. Contributions to compression standards like HEVC or its successor VVC, scheduled to be released in Q4 2020, aims at reducing these requirements, but such bitrate values are still hard to reach in a broadcast context.

Download the paper below