Simplify transparency handling by flattening image with white
background for bounds detection while preserving original
transparency in the final crop. This fixes the issue where
transparent areas were being treated as black during trimming.
- Replace OpenCV-based trim_whitespace with PIL implementation
- Remove redundant trim_whitespace_pil function
- Add initial trim step to extract command
- Standardize image processing to 2048px wide grayscale using Lanczos
- Handle transparency in whitespace trimming
- Use IntEnum for optimization levels
- Add proper level comparisons
- Treat transparent pixels as white for boundary detection
- Preserve transparency in output
- Fix black artifacts in transparent areas
- Add project description and features
- Include installation instructions with uv sync
- Add usage examples for all commands
- Document how each processing step works
- Add package installation instructions for different distros
- Modified deskew function to better detect staff lines
- Added morphological operations to enhance horizontal lines
- Used probabilistic Hough transform for more precise line detection
- Added .png to gitignore
Features:
- Page deskewing using contour detection
- Image enhancement with edge preservation
- High-resolution output (min 2048px width)
- Black and white conversion with proper contrast
Esa Kataja