Zero response so far over on DPReview so I'll present my new method of analysis here.
First a plea - I'm really interested to get hold of some long exposure darks from other Nikon cameras. Ideally around ISO 800 for 3-5minutes at room temperature. Noise reduction, long exposure noise reduction and high ISO noise reduction must be switched off. Send me a PM if you can help.
Now back to the analysis ...
It has been long established that the Nikon algorithm alters the value of an outlier pixel by using the values of its neighbours - typically the maximum value of those neighbours is used. My latest diagnostic method is to create a 2D plot of pixel value against the maximum of its neighbours for all the millions of pixels in the dark frame. In the absence of spatial filtering, here is the kind of result we would expect to find:
However, for the Nikon D5300 we find this instead:
What is going on? There is a very distinct boundary that has chopped off a complete "arm" of the plot, compared with the Canon. The boundary is easy to determine, so I've added a pink line to show it clearly. This pink line has a slope of precisely 2. In addition there is a very strong alignment of pixel values along a line with slope=1. In fact 5% of all pixels in the dark frame sit on that line with slope=1 i.e. the pixel value matches the maximum value of its neighbours.
There are no remaining pixels whose value exceeds twice the maximum of its neighbours (using a bias level of 588). It's clear that those outlier pixels have had their value truncated to match the maximum of its neighbours. This is the algorithm I described earlier in this thread.
Now let's look at the Nikon D7500:
In this chart I've used a heat map to count the pixels occupying the same position on the 2D plot:
Black represents count < 10
Green represents 10 <= count < 100
Yellow represents 100 <= count < 1000
Yellow represents 1000 <= count < 10000
White represents 10000 <= count
This looks quite different from the D5300. The pink line again has slope=2 and goes through the black level of 400. But the alignment of pixel values along a line with slope=1 has disappeared. Also the pink line is no longer a sharp cut off but it has still clearly removed a complete "arm" of the plot, compared with the Canon.
What I think is happening here is that the pink line with slope=2 does still represent a sharp cut off and that all pixels above that line (i.e. those with outlier values) have their value truncated. But instead of replacing the value with the maximum of the neighbours the value is replaced with a new value just above the pink line. With the D5300, 5% of pixels ended up on that line with slope=1 but with the D7500 statistical analysis shows that those 5% of pixels end up just above the pink line.
My hypothesis can be explained by an example. Take a pixel whose original value exceed the pink boundary line by 160. I think its replacement value is 160/8 above the pink boundary. A factor of 16 or some other factor might be used instead - it's very difficult to determine In any case, all pixels whose value originally fell above that line are pushed towards that line by the algorithm.
Now let's look at the Nikon 810A:
This plot looks different again and I haven't quite worked out what is going on. One thing that's certain is that each pixel is being compared to the maximum of 24 neighbours. One odd thing about the D810A is that in the raw file it is clear that very different digital scaling has been applied to each channel. It's possible this might have a bearing. I'll do some more channel by channel analyses.
Just out of interest, for comparison purposes, here is a 2D chart of pixels from a Sony A7RIII astro-image:
Note that the line with slope=1 forms a very strong cut off with very few exceptions violating this "rule". Statistics show a full 10% of all the pixels in that astro-image fall precisely on that line. Clear evidence of a quite destructive filter.
Back to Nikon - what about the effect of the spatial filtering on Nikon star colours? It certainly looks like the D7500 spatial filtering will have much less effect on star colours than the D5300. We have also determined that stars are almost completely unaffected by whatever spatial filtering the D810A performs.
So if you are worried about star colours, initial results are beginning to show that some Nikons are definitely better than others.
Mark
P.S. I'm really interested to get hold of some long exposure darks from other Nikon cameras
Edited by sharkmelley, 27 October 2018 - 06:12 AM.
