On the mDCT-PSNR image quality index

Full reference image quality indices assign a quality index to a pair of an undistorted reference image and a distorted image to be assessed; the quality of the index itself is then defined by its ability to predict the outcome of subjective tests performed by human observers judging the quality of the same image pair. In this article, a new DCT based image quality index is introduced whose complexity is between that of algorithms like SSIM and complex, HVS-based algorithms like VDP. Detailed experiments on the LIVE database show that the proposed algorithm performs best close to or below the visual threshold, and outperforms there existing algorithms like SSIM or VDP requiring only a mediocre complexity.