DocView Capture
Some parts of a page read badly however good the OCR engine is: a name written one letter per box, a faint carbon copy, a dot-matrix printout, a value on a ruled line. An image clean-up profile is a recipe of clean-up operations, in order, that prepares such a part before it is read. Finders read their zone (or whole page) through one or more profiles, and the Image enhancement step can apply a profile to every page.
Open Batch Settings › Image clean-up profiles.
Ready-made profiles
A new library starts with ready-made profiles for common situations. They are ordinary profiles: change them, switch them off or copy their idea into your own. Ready-made profiles added by a later version appear in your library once by themselves; Add the ready-made profiles puts back any that are missing.
The list has a Find box and pages, so it stays quick however many profiles you have.
| Profile | For |
|---|---|
| Character boxes printed in black (bold lines) | Forms with bold black boxes anywhere on the page, such as application forms: the boxes go and the handwriting stays, even where it touches a box. For whole pages or a zone |
| Names in character boxes (one letter per box) | First names, last names and initials written one capital letter per box, as on application forms |
| Character boxes printed in red / green / blue | The same, when the boxes are printed in a colour |
| Dates in boxes (DD MM YYYY) | Dates written one digit per box |
| Comb fields (open boxes with short teeth) | Account numbers, postcodes and codes written over a comb line |
| Handwriting on ruled lines · Underlined values | Text written or printed on a line |
| Table cells (remove grid lines) | Values inside a grid, so cell borders are not read as 1, l or | |
| Coloured form, red / green / blue | Forms printed in a colour: only what was written stays |
| Faint or light text · Light pencil | Pale print, faded copies, pencil |
| Bold, heavy or bleeding ink | Letters that run together |
| Dot-matrix print · Typewriter text | Older printouts |
| Carbon copy or NCR paper · Faded thermal receipt | Copies and receipts |
| Noisy fax or photocopy · Dirty paper and heavy speckle | Specks and grain |
| Shaded or coloured background · Stamps and highlighter over text | Text on colour |
| White text on a dark background | Inverted labels and headings |
| Photographed document · Low resolution scan | Phone photos and 150–200 DPI scans |
| Small print · Single characters and tiny zones | References, footnotes, single characters and ticks |
| Handwritten text (general) · Cheque MICR line | Free handwriting; the code line of a cheque |
| Scanner edges and punch holes · Standard office documents | Whole pages |
Making or changing a profile
- Click New profile (or a profile on the left), give it a name and say what it is for.
- Add an operation and set it up; add more and put them in order with the arrows. Each can be switched off without removing it.
- Save.
Besides the image enhancement operations (deskew, despeckle, black and white, line removal, colour dropout, contrast, sharpen and so on), profiles can use operations made for small zones:
- Remove character boxes on the whole page (bold lines): needs no zone. Removes printed box lines (long thin lines, at least Box lines are at least mm) and box walls (uprights of about a box's height that join two box lines beside a gap between boxes), keeping letters that touch or cross a box. A letter such as I touching both lines is kept too; switch on Boxes share their walls for grids without gaps between boxes. Raise Lines are at most for very bold boxes.
- Remove character boxes (one letter per box): removes box walls that run (nearly) the full height of the zone and box lines that run its width, keeping the letters, even where a stroke crosses a line. Lower Box walls are at least for comb fields whose teeth are short.
- Enlarge or shrink: OCR engines like text about 30 pixels high, so small text and small zones read better enlarged (or set a resolution such as 300 DPI).
- Add a white margin: characters touching the edge of a tight zone are often missed.
- Shape strokes: join dot-matrix dots and broken strokes, remove dots and fine noise, fill holes in characters, thicken or thin strokes (optionally only horizontally or vertically).
Testing before and after
Under Test before and after, choose one of a Batch Setup's samples (or upload a file) and a page.
- Draw a zone on the page with the mouse to test only that part, exactly as a finder reading through the profile does: the zone is cut out with a small margin, cleaned and read on its own. Without a zone the whole page is tested.
- Start from the cleaned page tests on the page after the Batch Setup's own image enhancement (what finders see); otherwise the sample's original file is used.
- With Read with OCR, what OCR reads before and after is shown with its confidence, so you can see whether the profile helps.
- Each operation's Test button tries that operation alone. The table lists how long each operation took and what it did.
Using profiles in finders
In a finder's settings (Batch Setup › Document classes › Finders), under Image clean-up before reading, add one or more profiles; they are applied in that order. Choose what is cleaned:
- only the zone it reads (zone finders: zonal, OMR, advanced zones): the zone is cut out, cleaned and read again on its own; the words are put back where they are on the page.
- the whole page: each page is cleaned and read again before the finder searches it (slower; finders using the same profiles share the work).
Before and after shows the finder's zone (or the page) on the chosen sample before and after the clean-up, with what OCR reads.
Using a profile for whole pages
In the Batch Setup's Image enhancement steps, Add a clean-up profile adds a step that applies a profile's operations to every page, in its place in the list.
Batch Setups keep their own copy
When a finder or an Image enhancement step is saved, the Batch Setup keeps a copy of the profile's operations. Changing or deleting a profile in the library does not change a released Batch Setup: save the finders again (or click Copy again from the library on the enhancement step) in a draft to take the change.