Determine the best event statistic under a two-pole pixel front end

Determine whether the first-interval statistic remains the strongest of the three candidate statistics under a two-pole event-vision sensor front end, or whether the rate statistic overtakes it.

Background

The paper’s analytical treatment assumes an event-vision sensor pixel with a single dominant front-end pole. Under this assumption, the linear-ramp analysis treats the pole primarily as a delay that cancels between consecutive events, making the first inter-event interval particularly effective for recovering the optical weight and reconstructing the modulation transfer function.

A more realistic second-order front end would affect the early event intervals because the transient response contains the second pole. The authors state that this additional pole would break the cancellation used in the single-pole derivation and introduce a term into the first-interval inverse that does not disappear under zero-frequency normalization. The unresolved issue is therefore which statistic is most robust or informative when the pixel model includes two poles: the first interval or the averaged event-rate statistic.