What steps will reproduce the problem?
1. analyze a light background image with Object Statistics
2. analyze a dark background image with Object Statistics
3. Compare results, in one case, the background pixels are considered
to be the only object in the image.
What is the expected output? What do you see instead?
Expected: Logic knows what background and foreground is, by manual
selection or analysis. Should not interpret background as
objects.
Observed: By default, only dark background images are correctly
analyzed. Light background images are considered to be made
up of one object, the background, due to comparison of pixel
value to threshold value ie: "if(value > threshold)" .
Please use labels and text to provide additional information.
Issue discovered while developing ObjectStatistics3D, testing on simple white field images. Not yet known how many other feature extractors have logic affected by this issue. It is necessary to study their operation and determine what threshold criteria are being used for object extraction.
Short term fix: All feature extractors should by default prefer to treat
images as black field.
Long term fix: A parameter should be built into affected feature
extractors allowing for bright/dark field selection accessible from
the GUI.
Original issue reported on code.google.com by Jon.Sand...@gmail.com on 14 May 2013 at 4:40
Original issue reported on code.google.com by
Jon.Sand...@gmail.com
on 14 May 2013 at 4:40