Adaptive Frequency Weighting for High-Performance Video Coding

Abstract

The benefits of utilizing the spatial-frequency sensitivity of the human visual system in video coding have been known for decades and are incorporated into video standards, such as MPEG2, MPEG4, and H.264/AVC, as well as into still image-coding standards, such as JPEG and JPEG-2000. However, all of these standards utilize frequency weighting that is adaptive only at the picture level. In contrast with such picture-level adaptive frequency-weighting algorithms, we describe a coding approach in which a macroblock-level adaptive frequency-weighting (MBAFW) algorithm is used. We show that for the bitrate, quality levels, and resolutions typically associated with high-definition video, MBAFW can lead to overall bitrate reductions typically in the 5%-9% range over the methods used in state-of-the-art video-coding standards.Index Terms-Frequency weighting, human visual system (HVS), perceptual coding, quantization matrix.

Advanced Video Coding Systems

“…In Tang et al (2006), the spatial-temporal sensitivity models are applied in rate control. In Chen et al (2012), adaptive quantization matrix selection and updating method is proposed, in which the macroblock is classified into three types: no frequency weighting, detailed preserving, and undetailed. In , the divisive normalization idea is applied in video coding, where the frame residuals are transformed into a perceptually uniform space by adjusting the quantization parameters for each MB.…”

Overview and performance analysis of AVS mobility features

“…According to the characteristics of Human Vision System (HVS) [14], the quantization quality of transform coefficients located at different positions inside a transform block can lead to different perceptual effect for human eyes [15]. In H.264/AVC, 8 x 8 and 4 x 4 non-uniform quantization weighting matrices are used to control the subjective quality for a whole picture.…”

“…Based on this observation, Adaptive Frequency Weighting (AFW) defines three type of weighting quantization matrices at the picture level and chooses one of the matrices at the macroblock level. Specifically, the three quantization matrices correspond to three sorts of quantization qualities, which are described as follows [15].…”

Weighting Quantization Matrices for HEVC/H.265-Coded RGB Videos

“…In [9], three types of macroblock-level WQMs are designed, each with seven levels of weighting factors at most. The WQMs are applied based on the characteristics of neighboring macroblocks to protect high frequency details.…”