Development of a digital compressed picture quality assessment system considering human visual perception

Abstract

At present, quality assessment of image codecs is performed exclusively by subjective assessment tests. However, these assessment tests not only are extremely labor and time consuming but also have limitations in that they do not allow real-time assessment. The authors have developed an image quality assessment system to enable accurate assessment of image codecs. This system involves computing the encoding error values between the synchronized original image and the encoded/reproduced image, weighting them by visual perception characteristics, and outputting the quality assessment values in real time. The authors were able to obtain a significant improvement of assessment accuracy by performing another weighting with consideration of the observation point scattering on the screen, which was used as a visual perception characteristic, in addition to such characteristics as the visual perception sensitivity in the local spatial frequency direction, and the noise masking effect. It was confirmed by comparison with subjective assessment tests that the current assessment system allows practically problem-free quality assessment. © 2000 Scripta Technica, Syst Comp Jpn, 31(5): 97–109, 2000