Problem ;
In scanner software there can be many different settings to adjust, to get the best result.
Each adjustment may need some playing with, and then a test scan done to check the result, or have look at an estimated preview.
This can take quite a bit of time.
Objective ;
The user is usually trying to optimise results in terms of parameters like ;
1) Clarity (sharpness, contrast, etc)
2) Colour Balance
3) OCR success rate
4) Minimal spelling mistakes in OCR
Solution ;
These (performance) parameters can be measured.
Thumb Nailing tests (like in photography) are a handy way to preview the results of a range of settings and compare them. They can also be used to compare the variation of one setting (horizontally on the page/preview display), while also varying another setting (vertically).
Previews may be calculated as "estimates" of what the result "may" look like. But thumbnails could be "done for real", using different settings for different parts of the image, or for the same small representative part of the image.
The user could choose whether they want to use the preview estimates or scanned thumbnail results for testing.
The user could click on the preview estimates or scanned thumbnail results to choose which one gives the best result. This is a bit like aligning the inkjet cartridges on most printers.
The user could also select the other results which are acceptable, which would help the software "learn" what the range of guesses should be in any future (or next) guesses.
The program could also ask the user to identify preview estimates that were inaccurate when compared to their actual resulting thumbnail - or the software could maybe make those matches itself.
For OCR, the spell checker can be used to check the OCR of the letters, and settings varied to give the best results using the spell checker. In areas of the page where OCR is needed, results may be improved by blurring rather than sharpening, and letter detection based on min and max line/curve width.
After a while, it might be possible to insert a page, the scanner does a high res scan, detects the quality of media, and knowing the objective of the scan (picture, text, or both) would detect the optimum settings, and set the scan to the optimum settings (option to ask the user if they are happy).
This would result in pictorial scans that would need minimal(if any) post scan processing (with loss of quality and wasted time).
This would result in OCR scans with very high recognition rates compared to other systems.
The optimum settings, and ranges for best guesses are useful data that could be sent back to the Ubuntu or Scanner software website, so these can be made available for statistical comparison, incorporation as optional default settings, and download by other users.
