Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Feature Request/Discussion: Add Otsu to Background Removal algorithm options? #1213

Open
kyrieb-ekat opened this issue Oct 1, 2024 · 0 comments

Comments

@kyrieb-ekat
Copy link

kyrieb-ekat commented Oct 1, 2024

The issues we've been encountering in Background Removal primarily has to do with the face it likes to scoop out the middle of neumes, and does so inconsistently in shape. This makes downstream classification in the IC and NIC difficult. In experimenting with other processing of images I realised that this kind of edge sensitivity but innards scooping was pretty standard for the Sauvola algorithm it defaults to.

The SAE binarization algorithm it has as the other option does a little better, but in general I've found that Otsu, while more sensitive than Sauvola with text and page shadows, does very nicely with grabbing a 'whole' neume. Given we can manually add and adjust text later in Neon, it might be worth trading okay text and bad neumes for terrible text and good neumes, maybe?

This might just be part of experimenting with the BR job and seeing what works (I've been tinkering with settings off and on, but the results with it don't beat what we can do by skipping the BR job), but I'd be curious to hear other folks' thoughts on adding this. Gamera4 has Otsu already, which I believe Rodan also technically has, so hopefully this wouldn't be too tricky...

Screenshot 2024-10-01 at 2 01 39 PM Screenshot 2024-10-01 at 2 01 45 PM

Sauvola, top; Otsu, bottom

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

No branches or pull requests

1 participant