Image enhancement and separation

Scanned pages are rarely perfect, and a stack of paper rarely arrives as neat, separate documents. The Batch Setup's Workflow tab fixes both: the Image enhancement step holds the enhancement steps, and the Separation tab of the OCR, separation & classification step holds the separation rules.

Image enhancement

Enhancement steps run on every page, in the order listed. The original scan is always kept, so enhancement never destroys what the operator captured.

Where it runs. Add the Image enhancement step to the workflow, right after Scan. It then enhances every page once, before the Image check, Quality check and OCR, so the check judges — and operators and OCR see — the enhanced pages. Without that step, the enhancement steps run inside OCR, separation & classification, as before. Pages replaced in Quality check after the Image enhancement step are enhanced by OCR.

Add an enhancement step

1. Open the Batch Setup, go to the Workflow tab and select the Image enhancement step. Create a draft first if the version is released. If the workflow has no such step, select OCR, separation & classification and click Add the Image enhancement step on its OCR tab (then Save workflow).
2. Under Image enhancement, pick a step in the list and click Add step.
3. Fill in its settings and click Save on that step.
4. Drag steps by the grip, or use the arrows, to change the order. Use the On switch to turn a step off without deleting it.

Try the result on a scanned page with Enhancement preview on the Scan screen.

Test the enhancement on a sample

At the top of the Image enhancement list, the Test box tries the steps on a page of your own without saving anything:

1. Choose a sample: a PDF, TIFF, JPEG or PNG, for example a crooked or speckled scan.
2. Click Test all steps to run every step that is switched on, in order, or click Test on one step to run that step alone.
3. The first pages (up to 4) show side by side, before and after, with their sizes and what the steps did, for example the angle that was corrected or Blank page. Click an image to open it at full size.

The settings on screen are used, even before you click Save, so you can change a value and test again straight away. The sample stays chosen until you leave the page.

Enhancement steps and their settings

Tip: A good order is: Auto-rotate, Deskew, Crop to content or Border removal, then tone steps (Clean background, Auto contrast, Drop out a colour), then Convert to black & white, then clean-up of the black and white page (Remove lines, Despeckle, Repair characters), and Blank page detection last. OCR, barcode reading and zones all work better on straight, upright, clean pages.

Note: Colour pages are kept in colour and grey pages grey, unless a step converts them. Black and white pages are stored as small grey PNG files.

Blank page detection counts real marks on the page and ignores the outer edge, dust and faint show-through from the back. Pages marked as blank can also start a new document with the Blank page separator rule below, even when you do not remove them in cleanup.

Document separation

Document separation decides where one document ends and the next begins. You add separation rules, and DocView applies them to the loose pages of each batch.

How rules work together

Rules vote on each page. The rule highest in the list that has an opinion about a page wins. A rule with no opinion leaves the decision to the rules below it. Drag rules, or use the arrows, to change their priority.

Without any rules, pages are not split automatically. Pages that operators already put into documents, for example in Scan, are never split again.

Add a separation rule

1. On the Workflow tab, select OCR, separation & classification and open its Separation tab. Under Document separation, pick a rule in the list and click Add rule.
2. Fill in its settings and click Save on that rule.
3. Order the rules and use the On switch as needed.
4. Click Test separation and classification with a sample » to check the result. See Testing a Batch Setup.

Separation rules and their settings

Each file is a document A new document starts with every imported file (upload, hot folder or e-mail attachment). Other rules can still split a file. This is the easiest choice when every PDF holds one invoice. No settings.

Fixed page count Starts a new document every N pages or sheets.

Blank page separator A blank page starts a new document. Remove the blank page is on by default, so the blank sheet does not end up in the output. This rule uses the pages found by the Blank page detection cleanup step, so add that step too.

Barcode separator A page with a matching barcode starts a new document.

For example, to split on barcodes that start with DOC-:

^DOC-

Patch code separator A Kodak patch code sheet starts a new document. Patch codes lists the codes that separate, one per line; DocView reads patch codes 1, 2, 3, 4, 6 and T. Leave it empty to use 2, 3 and T. Remove the patch sheet is on by default.

Separator sheet A printed separator sheet starts a new document and is always removed. Set Barcode value matches (a regular expression), Or text on the sheet (for example DOCUMENT SEPARATOR), or both. This rule needs OCR when you use text.

Keyword / regex separator A page with any of the keywords starts a new document.

Tip: Set Only in the top part of the page (%) to about 30 for invoices. Titles such as "INVOICE" sit at the top, while the same word further down the page (for example "invoice total") will not start a new document.

Page size and count A new document starts when the page size changes, or when a document has reached the number of pages set for its size. Use it for stacks of forms with other papers between them, for example 4-page A4 application forms with a cheque scanned after each one.

Under Page sizes, list one size per line as name: width x height = pages:

Form: 199-220x286-310 = 4
Small page: <199x<286 = 1

With the example above, every four A4 pages make one form, and each smaller page, such as a cheque, is a document of its own. Pages that match none of the sizes form an "other" group: they stay together until the size changes. The sizes are tried in the order listed, so put the narrower ones first when they overlap. In an exported Batch Setup file, a size can also be written as { "name": "Form", "size": "199-220x286-310", "pages": 4 }.

Classification change (advanced split) Pages are classified one by one, and a new document starts where the class changes. Tick Always let an operator check the split in Classification to send every batch to Classification, where operators can fix the split with drag and drop. Otherwise only batches with uncertain classes or loose pages go to Review. This rule needs good classification rules or a trained classifier.

Manual (operator splits) No automatic separation: pages stay loose and an operator splits them into documents in Classification or Index. When this rule is on, the other rules are not used. Make sure your workflow has a Classification or Index / Validate step, or switch on one-touch scanning (below) so operators split the documents while they scan.

One-touch scanning

One-touch scanning lets scan operators sort documents by hand as they feed them, much like OPEX-style manual sorting. The Scan screen shows a button for each document type. The operator picks up a document, presses its button (or key), keys any fields you ask for, and feeds the paper: a new document of that type starts and the scanner runs.

To set it up:

1. On the OCR step of the workflow, add the separation rule Manual (operator splits) and keep it switched on. Its card shows whether one-touch scanning is on, with a link to where you set it up.
2. Go to the Document classes & fields tab. The One-touch scanning: document type buttons panel is at the top. (It is greyed out until the Manual rule exists and is switched on, because the buttons are the document classes you set up on that tab.)
3. Switch on Use one-touch scanning in this Batch Setup.
4. Choose Buttons on the Left or Right of the Scan screen. Operators can still move the panel and change its width for themselves.
5. Leave Start the scanner when a button is pressed on (the default) if the scanner should start as soon as a type is chosen. Turn it off if operators prefer to press Scan themselves.
6. List the document types in the order you want, with Add a document type… or Add all. Use the arrows to reorder them. For each one:
   • Key: a shortcut from F1 to F12, or None.
   • Fields keyed before scanning: click + field… to add fields the operator types from the paper before scanning, for example a reference number. Required fields must be filled in.
7. Click Save one-touch scanning, then publish the version. See Versions and publishing.

Switching the Manual rule off or removing it switches the buttons off. The settings are kept while the rule is off, so switching it on again brings them back.

If you list no buttons, every document type gets one, in Batch Setup order, with keys F1, F2 and so on.

Documents started with a button keep the operator's document type: they are not separated or classified automatically, and the keyed field values are not overwritten by extraction. On a new batch, Scan asks the operator to choose a document type first. A type that was pressed but had nothing scanned after it is dropped when the batch is released. See Scan documents for the operator's side.

Common separation recipes

Coming from other products

If you used separator sheets, patch codes or barcode separation in Kofax/Tungsten Capture, IBM Datacap or ABBYY FlexiCapture, the same approach works here. Your existing sheets can usually stay in use: set the rule to match their barcode value or patch code.