No noticeable difference evaluation of image data compression

摘要

In this paper, a new way of evaluating image compression performance is proposed in order to reflect the human visual perception perspective in image compression and its applications: “No noticeable difference (NND) is allowed in commercial use.” Various methodologies have been proposed and utilized to evaluate image compression performance with both objective and subjective measures for the given rate-distortion curve, which is used to claim that a codec is better than another in a quantitative manner. In this paper, a NND-based subjective experiment is carried out to see if the existing standard image codecs such as JPEG and JPEG2000 still make a reasonable difference in performance. The experimental results on standard test images reveal that there is no great difference in performance between the two standard codecs including a few cases where JPEG outperformed JPEG2000.