Fidelity criteria for visual image transmission over noisy communication channels
Donald Edwin Pearson · Spiral (Imperial College London) · 1965
This thesis presents the results of a study of the problem of coding visual information for transmission to a human viewer.A model for a visual image transmission system is described which emphasizes the influence of the viewer on the parameters of the system.The model involves attaching a numerical measure of fidelity to any received picture and choosing codes which tend to maximize this fidelity.In this connexion reference is made to Shannonls work on the coding of sources with a fidelity criterion.The model is shown to elucidate certain problems in the bandwidth-compression of visual image signals, where the reduction of channel capacity and the preservation of adequate fidelity in the received pictures are conflicting requirements.In some experimental studies the problem of measuring the fidelity of a received picture is extensively examined, and some comparative trials of rating scales and pair-comparison methods are reported.method of fidelity measurement is suggested in which the unknown distortion and a reference standard are presented in temporal succession, using the sequential estimation technique of Dixon and Mood.By means of this method, the subjective cost of position errors resulting from channel noise in run-coded pictures is evaluated.It is shown that the cost of the errors may be represented by single-symbol distortion measures, allowing the calculation of a rate-distortion function R(D) for the source.Further experimental studies are reported of the effect of very high channel noise levels'in conventional television systems, and of the ability of human viewers to detect the presence of picture signals in the noise. W 2Position channel bandwidth.X Transmitter channel alphabet.X 1 (t) Luminance signal.X 2 (t) Position signal.x.A transmitted channel symbol.Y Receiver channel alphabet.Yj