A perceptual quality based rate distortion model

Rate and distortion models (R-D models) are important for current image/video encoders. They can help encoders to find the best encoding parameters for improving encoding performance quickly and effectively. However, most of the existing R-D models are based on mean square error (MSE). For the purpose of achieving better perceptual quality in encoding, the R-D models based on perceptual distortion measurement are involuntarily needed. In this paper, we proposed a perceptual quality based R-D model for lossy image/video encoding. The experimental results show that, the proposed model can accurately describe the R-D characteristics of the sources in the context of perceptual distortion, both for whole frames and the areas inside frame.