Pixel Math error

[Peter]

Hello everyone, following Gerald's excellent pixel  math videos and I am creating black, grey and white images as shown with the statistics to measure as per instruction. The black reads 0.000 on all measurements so that reads ok, the grey image reads average readings of 0.502? But the problem is with the statistical reading of the white image which gives 0.000, the same as the black image. If I reduce the intensity from 1 to 0.9 then the statistics read 0.900 as expected. If I got to 0.99 then the statistics revert to 0.000! Any ideas as to the reason why the whites read the same as the blacks? I use normalise 0/1, create new image, RGB or greyscale makes no difference to any of the readings of course. So any advice out there?

[oldwexi]

Hi Peter!
I thinks thats a bug in Statistics.
Created an image with "1"  in PixelMath. Got a white image, but
Statistics Displays 0, 0, 0 in all Areas
Readout Cursor Display 1, 1, 1 everywhere

Win 7, PI 1.8 - 1087

Gerald

[georg.viehoever]

I think that some tools ignore pixels very close to zero or 1 (because they contain no useful information). Maybe Statistics is one of them.

[Peter]

Thank Georg and Gerald I can always count on sound and explanatory comments from both of you, my thanks to you bo :Dth

[silentrunning]

I get this same problem.

I'm following Gerald's tutorial on PixelMath Basic Functions and statistics displays a pure white image as having pixel values of 0 rather than 1 though it works for him in his tutorial. Is this bug being addressed, as a beginner with PI it completely confused me and this thread was started sometime ago.

Many thanks

Doug

[oldwexi]

Hi Doug!
This bug is still existent.

But, if you enter in PixelMath instead of    1   the expression
1 - 0.00001
you get also a white image but Statistics now displays the values correctly

Using the ReadoutCursor shows in any case the correct values.
No need to be confused.

Gerald

[silentrunning]

Many thanks Gerald.

Any particular reason why it hasn't been fixed, seems like an easier enough thing to do. How come it works in your tutorial?

Doug

[oldwexi]

Hi,
assume in 1.7 it did not exist.
However it not an error in PixelMath PI 1.8
it is an issue in statistics PI 1.8.

Gerald

[kwiechen]

Hi,

the Statistics module seems to ignore 0 and 1 pixel values. However this is an annoying feature because the PJSR Image methods like Image.mean() do not ignore these pixels like Pixelmath. At least it should be possible to set some constraint (upper lower values) when computing statistics with PJSR and Pixelmath.

Kai


 

[oldwexi]

Kai!
I think it has nothing todo with PixelMath.
ANY process which produces a white Image creates 1.0 in the Pixel value (as also PixelMath does)
The readout Cursor than shows 1.0, so the data are o.k.  !

ONLY the STATISTICS Process in PI 1.8 displays the Pixel Values WRONG, despite the correct data!

You can work with PixelMath and other processes afterwards they calculate on the base of 1.0 very precise
and correct.

Gerald

[kwiechen]

Gerald,

sorry, this was a typo.

the Statistics module seems to ignore 0 and 1 pixel values. However this is an annoying feature because the PJSR Image methods like Image.mean() do not ignore these pixels like Pixelmath Statistics module. At least it should be possible to set some constraint (upper lower values) when computing statistics with PJSR and Pixelmath.

The Statistic module ignores 0 and 1 pixels. However all processes within PI should give the same result. So it would be nice, if for example the PJSR Image functions like Image.mean() and the Pixelmath mean() function get this capability too (with a simple boolean switch added for example)

Kai