I have got many hundreds of old and not-so-old technical PDF files with plenty of diagrams lying around and I know the pain of digitizing the curves and converting measurements back into the original units.
What I would like to see is a semi-automated workflow that could take as an input a cropped image, possibly ask me some questions or make me point-and-click at the picture and output (x,y) pairs for each curve (either raw data at each pixel, or post-processed splines).
- Any raster image format will do, PNG and JPEG will be fine, ability to accept raw PDF - outstanding
- CSV is OK.
- Ability to work with noise introduced during scanning
- Ability to cope with multiple curves of same or different colors and patterns
- Ability to process logarithmic axes
- Ability to process nomograms
- Portable (Windows, Linux)
Related: Extract numbers from a figure (to my mind, not a duplicate because of different requirement sets).