|
2. Jun 2013|00:41 |
Pie tam, cilvēka vizuālā uztvere nav īsti rotational-invariant.
One, the pattern-matching wetware in our brains takes a time to activate that is approximately proportional to the angle of rotation between the comparable samples. Two, our vertical edge detectors are typically more sensitive than our horizontal edge detectors. Three, some of the heavy first-order encoding and pre-processing work is done as early as in our retinal connectomes, to pass for the brain as little data as possible. |
|