Script with GUI that removes screentones, or the "printed" effect in manga and doujins, using DSP. This process is necessary for use with DeepCreamPy, by deeppomf, and my other project hent-AI. Gaussian blur is applied to an image, then an averaging blur afterwards. This removes high frequency signals (screentones). The output of this is sharpened with a Laplacian kernel, to retain some edge.
Here is a visual example: NOTE: More blurring will remove heavier screentones, but will result in more loss of quality from the image. You may need to experiment with the 3 blurring levels depending on the artists screentone usage.
You should have whatever doujin or images you want in .png format, in some folder. You should also create an output folder. I require .png because DeepCreamPy requires png format. nhentai.net may download jpg files, but you can batch convert using this jpg2png
Windows, tested on windows 10.
Python 3 if you want to run the source codes. Uses opencv (cv2) libs, numpy, and tkinter for the UI.
Note that the .exe file is quite large. The signal processing utilizes a large library of signal processing related functions, opencv.
-
Executable
Simply follow the link above and download the package.zip. Extract the program from the zip. I would reccommend creating an input folder and output folder in the same directory for ease of use. There is a slim chance that the file directory dialogue can trigger your Anti-virus. I was able to bypass this by just closing the app and restarting it again.
stremove.exe
- Source Code Simply run stremove.py. I use anaconda on my windows machine.
py stremove.py
You should use this in tandem with my other project, hent-AI, and DeepCreamPy, by deeppomf.
Any contributions are welcome, I am still a student with limited resources and time and appreciate any help I can get.
-
1.6: Added support for jpg files. But all images should be in png to ensure compatibility with DeepCreamPy.
-
1.5 Cut down on unneeded libs (file size unchanged), fixed bug where .PNG wasn't recognized
-
1.41: Greatly improved sharpening process, which was brightening images unecessarily due to a miscalculation. Updated labels and defaults.
-
1.4: Sharpening is now parameterized. You can customize the filter to suit whatever art style you encounter. Defaults are still visible. For more information on the filter params, check out this page
-
1.3: * Updated blurring process. Secondary blur is now a bilateral filter, which retains edges and removes toning much better.
- Removed secondary sharpening because it can add noise in some circumstances.
- Default slider position is now on 2.
-
1.2: Removed redundant modules. Created exe file.
-
1.1: GUI, directory selection, batch screentone removing, variable removal
-
1.0: Can remove screentones from an image
- Allow custom kernel types for blurring and sharpening
- Parameterize Bilateral Filtering inputs for Sigma Space
- Better UI
- Find optimal default values
- Remember and autofill Input/Output directories
- Somehow reduce file size
-
Nathan Cueto - Everything - My GitHub
-
Artist soranosuzume for the sample image.
This project is licensed under MIT