a7RII FW 4 star-eating test results

Started Jun 12, 2017 | Discussions thread
sharkmelley
sharkmelley Senior Member • Posts: 2,298
Different algorithm, same power spectrum?
8

Hi Jim,

Thanks for proving beyond doubt that the Star Eating behaviour still exists in A7Rii firmware version 4.00

However, I was initially puzzled that your power spectrum is showing no difference in the green channel even though I'm convinced that FW 4.00 treats the green channel (but not the red and blue channels) slightly differently to FW 3.30.

Here is a diagram that will help explain the difference between the original spatial filtering algorithm(FW 3.30) and what I think is the new spatial filtering algorithm (FW 4.00):

For both algorithms, if the green pixel V is brighter than its green neighbours then its value is replaced by the maximum value of its green neighbours. If instead the green pixel V is dimmer than its green neighbours then its value is replaced by the minimum value of its green neighbours. Otherwise the value V is left unchanged.

The difference between the algorithms is in which set of pixels are considered to be the neighbours. The old star eater would consider the 8 pixels V1, V2, V3, V4, V5, V6, V7, V8 to be the neighbours. The new algorithm considers the 8 pixels V5, V6, V7, V8, V9, V10, V11, V12 to be the neighbours.  At least this is my hypothesis from closely examining the data after the spatial filter has been applied.

Importantly, for the old star eater algorithm, the two green channels G1 and G2 could be processed independently.  But the new algorithm requires the green channels to be processed together because 4 neighbours come from G1 and 4 come from G2.

I did a quick test by taking some A7Rii raw data (a 3.2 sec dark-field exposure) and then applying either the old or the new spatial filtering algorithm to it.  I then took the Fourier transform of the same 1024x1024 block of one of the green channels and summed the columns - essentially producing the power spectrum:

I was unable to discriminate between the two algorithms by looking at the power spectrum.  In other words, it appears to me that the power spectrum is not sensitive enough to be able to distinguish between the 2 algorithms.

If I'm right, then this could explain why your power spectra remain essentially unchanged even though the spatial filtering algorithm may have changed.

Remember though, this is dark-field data which by nature is essentially random, so the difference is very subtle.  However, the same algorithm change when applied to a star field could result in more tiny stars being preserved.  In other words the effect of the algorithm change on random data might not be an accurate indicator of how well the algorithm behaves in a practical situation.

Mark

 sharkmelley's gear list:sharkmelley's gear list
Sony a7S Nikon Z6 +1 more
Post (hide subjects) Posted by
(unknown member)
(unknown member)
(unknown member)
(unknown member)
(unknown member)
(unknown member)
(unknown member)
(unknown member)
(unknown member)
(unknown member)
(unknown member)
(unknown member)
Keyboard shortcuts:
FForum PPrevious NNext WNext unread UUpvote SSubscribe RReply QQuote BBookmark MMy threads
Color scheme? Blue / Yellow