Image quality assessment based on pyramid decomposition and mean squared error

Widely used image quality metric peak signal-to-noise ratio, PSNR, doesn't correlate very well with human judgment. In this paper, we show that using multi-scale image decomposition before calculating mean squared errors at higher pyramid's scales, we get multi-scale MSPSNR metric that is very well correlated to subjective assessment and very efficient for computation. Proposed metric is tested using LIVE3D database which contains stereo images with common distortion types.