Methods Equal Areas Filter

ImageFilter(basis="block") -- the block-basis alias

Numeric online method. Source: streaming/filter.py. Invoke via ImageFilter(expr, var, p0=, window_size=, ..., basis="block") then flt.partial_fit(t, y) per sample; flt.predict(x), flt.params_.

ImageFilter(basis="block") is ImageFilter with basis fixed to "block" -- the streaming twin of batch EAC. Its window measurement is the block image: window sums against a diagonal Gram, the cheapest per-sample statistic of the two bases and the one an embedded target runs (see the embedded tool). Its sibling ImageFilter(basis="legendre") fixes the Legendre basis instead, whose spectral measurement resolves an oscillatory plant's shape and frequency directly.

Everything else -- the whitened window-image measurement, the information-form update, drift detection through the shared DriftDetector, the adaptive window, robust winsorization, result()'s calibrated covariance and coasting -- is ImageFilter's and is described once, on its method page.

Where it is best applied

Use ImageFilter(basis="block") for monotone or saturating plants where the block image's low per-sample cost matters, or as the streaming path an embedded target runs. For an oscillatory plant use ImageFilter(basis="legendre"). For an accurate static batch fit use LSI or EAC.