Adaptive Quantization Matrices for HD and UHD Display Resolutions in Scalable HEVC

TL;DR

This paper introduces an adaptive quantization matrix method for HEVC that considers display resolution, significantly reducing compression artifacts and improving video quality for HD and UHD displays.

Contribution

It proposes a novel adaptive quantization matrix technique based on human visual system principles, tailored to display resolution, enhancing HEVC video compression performance.

Findings

Up to 56.5% luma BD-Rate reduction in SHVC enhancement layer.

Encoding time reduced by approximately 1%.

Improved visual quality on HD and UHD displays.

Abstract

HEVC contains an option to enable custom quantization matrices, which are designed based on the Human Visual System and a 2D Contrast Sensitivity Function. Visual Display Units, capable of displaying video data at High Definition and Ultra HD display resolutions, are frequently utilized on a global scale. Video compression artifacts that are present due to high levels of quantization, which are typically inconspicuous in low display resolution environments, are clearly visible on HD and UHD video data and VDUs. The default QM technique in HEVC does not take into account the video data resolution, nor does it take into consideration the associated display resolution of a VDU to determine the appropriate levels of quantization required to reduce unwanted video compression artifacts. Based on this fact, we propose a novel, adaptive quantization matrix technique for the HEVC standard, including Scalable HEVC. Our technique, which is based on a refinement of the current HVS-CSF QM approach in HEVC, takes into consideration the display resolution of the target VDU for the purpose of minimizing video compression artifacts. In SHVC SHM 9.0, and compared with anchors, the proposed technique yields important quality and coding improvements for the Random Access configuration, with a maximum of 56.5% luma BD-Rate reductions in the enhancement layer. Furthermore, compared with the default QMs and the Sony QMs, our method yields encoding time reductions of 0.75% and 1.19%, respectively.