Statistics displays wrong values when the whole image is white (1.0)

[oldwexi]

Hi,
There is a bug in "Statistics".
WHATEVER process you use which creates a white image (where all Pixels are set to 1.0)
Statistics displays pixel values 0.0  instead of 1.0

Using the Readout Cursor it correctly displays the Pixel values 1.0

Gerald

[georg.viehoever]

The statistics process should probably have a toggle setting "Ignore saturated pixels or 0 pixels", instead of silently ignoring them.
Georg

[Juan Conejero]

Hi Gerald,

The values displayed are not wrong. The Statistics tool excludes pure black and white pixels by default, as these pixels carry no information on a representable image.

In the next version we have included an "unclipped" option to disable automatic clipping; see the attached screenshot. Of course this option is disabled by default, so 0/1 pixel samples continue being excluded by default.

[kwiechen]

Is it possible to exclude 0 and 1 pixels from PJSR Image.mean() and PixelMath's mean() functions and others too?

Best regards,

Kai

[Juan Conejero]

Is it possible to exclude 0 and 1 pixels from PJSR Image.mean() and PixelMath's mean() functions and others too?

Sure. Use the following properties:

Number Image.rangeClipHigh
Number Image.rangeClipLow
Boolean Image.rangeClippingEnabled

to define range clipping bounds and enable/disable range clipping for a particular image.

As for PixelMath, the next version distributed with core version 1.8.3 comes with a special symbol initialization function that allows you to access view properties such as Mean, Median, StdDev, MAD, etc, in both clipped and unclipped versions. As of version 1.8.2, only clipped statistics are available through the mean($T), median($T), etc.

[kwiechen]

Juan,

thank you for the info. This is a handy feature to obtain information about a pixel subset in the image.

Best regards,

Kai 

[kwiechen]

PS this works nicely! Can't wait for more PJSR documentation and integrated C++ support .......

Best regards,

Kai