Error visualisation and presentation in video compression systems

As part of a program of work to explore video compression, the author has studied various aspects of human visual perception and modelling. Tests have been devised to evaluate visual acuity to dynamic effects on both interlaced and progressively scanned video. The results of these tests indicate that the visual sensitivity to dynamic errors is substantially different to static. A further test was devised to explore the aspect of error masking in visual perception. Such masking is widely known in audio coding and also apparent in the human visual system (HVS). The paper also describes an error weighting mechanism. Based on measurements of the reconstruction error in a compression codec. The effect is most noticeable in systems using high compression ratios. A compensation technique is applied to take advantage of the error weighting with the result of improved picture coding quality. The final part of the paper describes new work which attempts to measure the HVS characteristic in conjunction with the entropy coding efficiency. The tests reported attempt to find the optimum quantiser profile from the matching of entropy coding efficiency against HVS response