How do I add a review feature that highlights markers based on very small measurements?
Hi,
I am using the Otolith app to measure fish structures. When I get to the "Circuli Marks" page, and have hit "Find Circuli Marks" and then review circuli marks by adding and removing extra markers, there are occasionally extra very small makers that are difficult to see. These are markers less than 0.006 mm, ~4 pixels or less distant from previous marker, as measured along the transect.
These markers are usually erroneously placed after the "Find Circuli Marks." I realize that changing Smoothing and Threshold settings would help, however, there is a production process for the lab I work in so unfortunately there would be no single setting that works for the wide variety if image qualities (focus, lightness, shadings) present for the thousands of structure images and users almost never change these smoothing and threshold settings. Also, these settings work most of the time, I am wanting to catch the rare 1 to 5% of images on which there are extra meaningless circuli markers that are very small.
Made up example of very small circuli:
Would it be possible to highlight circuli markers, as a differnet color or some other highlight, on the "Circuli Marks" page that are less than so many pixels or measurement on the Circuli Marks page of the Otolith app? Would you recommend highlighting these after the "Find Circuli Marks", or any time that they exist on the transect?
Thank you.
Answers
Have you tried the Rising or Falling edge detection algorithms? In some cases they could work better. Also, you could use Navigation feature to verify Circuli marks.
For future investigation, could you post a sample where erroneously placed after the "Find Circuli Marks."? Please post original image and the measurements .lps file (save measurements on the app's last page).
Thanks,
Nikita.
Hi NikitaRG,
Yes, we use the Raising edge detection algorithm. "Raising" always nearly always works better as what we are marking is generally where there is a transition along the transect from dark to light. Every time circuli markers are placed, they are reviewed, so yes, we are using the Navigation feature to verify circuli marks. Users magnify the image to an appropriate level while they are reviewing markers. This process works great most of the time.
An additional feature that would be helpful for the rare occasion that markers are found very close together, is to highlight markers that are unreasonably (for our purposes) close together. These are missed because visually, markers that are very close together can be difficult to review. Here is an portion of an example image (these circuli are ~8 pixels, or 0.01 mm, I would want to set and have the ability to change the maximum measurement/pixel width highlight level):
When I try to attach the .lps, I get a "file is not allowed" error, but I have attached the snapshot/overlay and the original image for now...
Thank you,
liwilson
Edge detection tool has several parameters that can be used to eliminate false detection, for example increasing the Smoothing will detect edges that are more far apart.
Below is the screenshot with parameters on the same image (I used Peaks with Smoothing 14):
Some week edges might be missed, but they can be added later manually.
Yuri