Adaptive quantization for perceptual video coding

摘要

In one embodiment, a perceptual video coder obtains local video content, and calculates a feature index for each of one or more features of a video unit of the local video content. Each feature index is then directly mapped to a delta quantization parameter (QP) using a linear fractional model that explicitly establishes a relation between the delta QP and a Human Visual System (HVS) insensitivity index, e.g., where the delta QP is set based on a comparison between a local strength of the HVS insensitivity index and an overall strength of a particular frame of the local video content. Accordingly, a QP of an adaptive quantization may be refined by adding to it each delta QP, such that adaptive quantization may be performed on the video unit based on the refined QP.