Scott Manley

Of course that’s your contention. You think you’re clever because you can recite Ulaby’s textbook definition of an impulse response. Next year you’ll read Schwartzkopf’s, “Radar Generalized Image Quality Equation Applied to Capella Open Dataset,” and for a while you will be convinced a Shannon-theoretic approach to defining SAR “quality” is the “right” way to think about it. Well, what happens when I multilook ten times? Did that make the resulting image ten times more interpretable? Is interpretability even what you are proxying? Why should a 25cm single look with a -12dB NESZ at the scene center rate more highly than a -32dB 1m spotlight? Is it more or less interpretable? Next you’re going to pick up Doerry’s “Notes on Synthetic Aperture Radar Image Quality,” and realize it’s a lot more complicated than you thought…you’re walking down a dark path, and I don’t think you’re going to like where it leads. Is a NIIRS-scale-inspired statistical regression fit to a curve set by an inter-rater reliability study really the best we can do to reduce quality to a number? Why should human labelers be treated as a gold standard, when we already know most of the information is hidden in phase space beyond their meager faculties to perceive? And as we enter the dawn of machine super-intelligence, where swarms of PhD signals experts with organs that natively parse raw compensated phase history can be invoked with the slip of a mouse, should we not recall Fienup’s work on detecting movers by phase-error correction and sharpness gain, which showed definitively that movers apparent in phase space lay hidden below the clutter floor in associated detected products due to smearing in azimuth? Should we not expect, as in chess or Go or now the entirety of mathematics, that these machine intelligences will develop techniques so beyond our comprehension that we’ll be reduced to asking them to verify each other’s conclusions on our behalf, and that eventually we will stop asking even for explanation? What does quality mean in that future! What does an image mean in that future! How good is good enough!