The original version of your algorithm is:
For each pixel used,
1. Take two successive measurements
2. If the first measured brightness < the second measured brightness, output 1
3. If the first measured brightness > the second measured brightness, output 0
4. If the first measured brightness = the second measured brightness, do nothing
The approach is largely inspired by von Neumann’s randomness extraction algorithm:
1. Take two successive bits
2. If the first bit is 0 and the second bit is 1, output 1
3. If the first bit is 1 and the second bit is 0, output 0
4. If the first bit is equal to the second bit, do nothing
It wastes a lot of entropy. But its intended use in psi experiments influenced the design. Pixel brightness is proportional to the raw analog signal produced by the photodiode, which is proportional to the number of detected photons. Since it is hypothesized that post-processing can reduce the effect of psi, it aims to extract randomness as close to the underlying uncertainty or changes in the raw photon count as possible.
Your randomness extraction algorithm is more efficient and should produce random bits faster than the original version, but it may not align with the original design philosophy.
That said, whether the original version performs better in practice is unproven.
Anyway, for the experiment I’m trying to replicate in this thread, no randomness extraction algorithm is needed. The main focus is analyzing the unprocessed/unextracted noise properties such as autocorrelation and spatial coherence, rather than generating random bits.