image-analysis-training-resources issueshttps://git.embl.de/grp-bio-it/image-analysis-training-resources/-/issues2019-06-10T21:04:35Zhttps://git.embl.de/grp-bio-it/image-analysis-training-resources/-/issues/32Binarization exercise2019-06-10T21:04:35ZChristian TischerBinarization exercise@halavaty @hossain
I am struggling to find a binarization exercise that has the right level.
Either I feel they are too easy or I feel they sneakily require new concepts.
This is my current attempt (https://git.embl.de/grp-bio-it/ima...@halavaty @hossain
I am struggling to find a binarization exercise that has the right level.
Either I feel they are too easy or I feel they sneakily require new concepts.
This is my current attempt (https://git.embl.de/grp-bio-it/image-analysis-training/resources/blob/binarizationExercise/modules/binarization.md):
- Open image: "xy_8bit__binarization_exercise.tif".
- How many 'signficiantly different' (by more than 10 gray values) threshold values can you find such that there is exactly one connected foreground region?
<details>
<summary>Solution</summary>
There are four such threshold values, e.g. 8, 46, 94, and 111. There are more, but they are within 10 gray values of those four and thus, according to the rules of this exercise, do not count as significantly different.
</details>
<br>
<br>
The issue could be that one kind of needs the concept of a "connected" region (possibly raising the 4- vs 8-connected issue) and it also somewhat introduces a "noise" concept. But maybe that's fine, or even good because it would motivate the "Learn next" modules? I am not sure. Maybe it is not appropriate and one should find something easier...?!