Perceptual video coding with H.264

In this paper a novel approach to perceptual video coding is presented for block-based video coders. This is done by replacing purely mathematical models for measuring the distortion in video quality, such as Mean Squared Error (MSE) or Mean Absolute Difference (MAD), with a distortion model that measures the perceived distortion by an average human observer. Additionally, to keep the computational complexity of perceptual measurements reasonably low, the suggested model uses the structural properties of the underlying video coder to calculate the suggested perceptual distortion. It's shown that the distortion measure based on the suggested model can be utilized for different rate-distortion optimized video coding techniques such as mode selection and rate control.