Capture
Capture covers everything that turns a paper or electronic source into an indexed, usable image or record. The exam tests your ability to pick the right input path, scanner settings, recognition technology and index method for a business case, and to spot settings that quietly destroy quality or inflate storage.
Input paths and scanner choices
- Centralized production scanning with a sheet-fed ADF suits high-volume, uniform batches; flatbed or overhead scanners handle bound, fragile or oversized originals.
- Distributed capture at an MFP or desktop pushes indexing to the point of origin; mobile capture is convenient but weak on lighting, skew and resolution.
- Electronic sources (fax servers, email, e-forms, EDI and print streams, report archives) should enter the same repository with the same index structure as scanned pages.
Image settings that matter
- Resolution: 200 to 300 dpi handles clean bitonal text; 300 dpi is the usual floor for OCR, ICR and small-font forms. Past roughly 400 dpi you add file size with little accuracy gain.
- Bit depth: 1-bit bitonal (CCITT Group 4), 8-bit grayscale, 24-bit color. Color serves human readability, not OCR.
- Enhancement chain: auto-crop and deskew, border and punch-hole removal, despeckle, blank-page removal, auto-orientation, halftone removal, and adaptive rather than fixed thresholding on uneven backgrounds.
- Compression is applied at scan and again at storage: G4 for bitonal, JPEG for color and grayscale, JPEG 2000 or PDF containers where offered.
Recognition and indexing
OCR reads machine print, ICR reads handprint, OMR reads marks, barcode readers decode symbols. Index values come from manual keying, barcode-to-database lookups, zone OCR, or full-text extraction. Validate with check digits, date ranges and lookup tables, and route low-confidence fields to second-pass verification.
Worked example
A 12,000-page monthly invoice batch mixes black text with color logos, must be full-text searchable, and carries a routing barcode. Reasoning: scan at 300 dpi color so logos render correctly and the barcode reads reliably; store a hidden OCR text layer in a PDF container; populate vendor and PO fields by barcode lookup rather than keying. Size check: 8.5 by 11 inches at 300 dpi is 2,550 by 3,300 pixels; 24-bit raw is about 25 MB, JPEG at 10:1 about 2.5 MB, so roughly 30 GB a month. Bitonal G4 would cut each page to about 50 KB, but you would lose the color the business asked for.
Traps that cost marks
Confusing dpi with bit depth in sizing questions. Assuming more dpi is always better. Letting scanner-side JPEG compression degrade text before OCR. Keying a field that a barcode could populate more accurately. Forgetting blank-page removal and batch reconciliation counts.