SOFTWARE DREAMS
SCANNING AND OCR TECHNOLOGIES

GETTING THE MOST FROM YOUR SCANER OCR. In general older [386/DX PCs] Optical Character Recognition Programs [OCRs] run faster with 486/DX2+ [and higher] computers. However, don't be fooled by "horse power". Faster processing does not mean greater accuracy. Chip power alone, is not enough; and optical character processing is no horse race.

GREATER OCR ACCURACY REQUIRES GREATER PRINT TEXT QUALITY. The greatest diversity and poorest quality examples of print text can be found in newspapers and many government documents.

OCR ACCURACY IS BEST, WHEN THE OCR WORKS LEAST. OCRs do their best work with flat, clean, clear, non- glossy, plain paper, monotone, uniform print text.

While some OCRs work faster, processing speed is not directly related to accuracy. Except to note that poor print quality, slows processing time, while decreasing accuracy. Regardless of brand, as OCRs work harder ,accuracy declines. The more choices the OCR must make ... the more chances for an OCR mistake.

"SMART" COPIERS CAN IMPROVE THE GRAPHIC QUALITY OF PRINT TEXT. Most commercial plain paper copiers have graphics quality controls. These machines can be found in your local neighborhood Print Shop. Look for a print shop that keeps a copier "behind- the- counter". Public accessible and "coin-op" machines are usually over- used and problems go unattended.

Once you find a location, make sure the machine is well maintained and in good repair. An over- used, over- worked, copier is worse that having no copier at all. Most fully functional commercial copiers with graphics control can produce a copy that is "better" than an original with degraded text. To test a copier use a "spellchecked" original with "fuzzy" text. The new copy should be clean of speckels with clear text.

1. Clean and clear text background.
Remove "spots" and restore a clean background. Remove "color" backgrounds.

2. Clean and clear text print characters.
Remove fuzzy [print- bleed] edges from print. Yes, if the first copy shows improvement, a copy-of- a- copy may improve it more. However, re-copying from copies ... only works twice before visible improvement can no longer be seen.

3. Increase or decrease uniform text type size.
Make varied text the same scale. OCRs work better with text that "looks the same".

4. Reprint slick- gloss paper to plain paper.
Slick paper does not "feed" right. Glossy paper "glare" does not "read" right.

5. Cut and paste multiple print text columns to one continuous text column layout.
Avoid OCR decolumnization slow- down and errors.

6. Use a "fresnel" anti-glare plastic copy sleave to hold text and reduce edge shadows. Use only to copy and store in protective envelope or folder to avoid scratches. Ask for Multi Pad at local Office Supply Stores.