It's all Backwards: Rethinking Frame Rate and Temporal Fidelity in a Cinema Workflow

摘要

In pursuit of greater fidelity in motion picture images, it is important to carefully consider the frame rate, and particularly the way in which it is handled from acquisition to final exhibition. In typical workflows, the frame rate is decided prior to acquisition, and that frame rate is carried through editing to exhibition without change. Then, when slightly different frame rates are needed for international presentation or on devices with a fixed frame rate, complex algorithms are employed to make the slight increase or decrease in the frame rate from the fully finished product. This often results in substantial degradation to the temporal domain. The first question addressed is: why is frame rate conversion so difficult? Without very complex algorithms, slight changes to the frame rate result in unacceptable motion distortion. Because of the enormous level of temporal aliasing introduced by conventional cameras, it's not feasible to use one-dimensional resampling algorithms, and instead highly nonlinear "optical flow" algorithms must be employed. If, however, steps were taken to substantially reduce the temporal aliasing of the capture system, the frame rate conversion problem would be reduced to a straightforward one-dimensional resample with nearly no distortion. This result then leads to the next conclusion: for best temporal fidelity, frame rates should be as high as possible throughout production, with the aesthetic look controlled by adjusting the frequency response within that frame rate. Then the final output frame rates for any deliverable is created from that high-bandwidth master, in exactly the same way spatial resolution is currently handled.