OCR software will not output an error free document. Organizations usually need to correct spelling errors and fix the page layout. Scanned documents often contain imperfections, such as speckles, dots, and black borders. You will be amazed to discover how you improve the results, when you apply the following techniques:
• Scanning Resolution: Object Character Recognition works the best with documents scanned at 300dpi resolution. Scanning at higher resolutions and color scanning significantly increase the time it takes to scan a document. Increasing the resolution to 400dpi for spreadsheets, ledgers and old newspapers may improve the results. Typically however, you do not need to scan at 600dpi unless the pages have very small font sizes less than 6pt.
• Color Scanning: Deteriorated documents may also be completely unreadable when scanned in black and white (B&W) mode. Color and grayscale scanning improves the recognition rates for old documents that are yellowed, stained, wrinkled, and faded. Color capture may also improve the recognition rates of documents with color backgrounds, shading or small fonts and line breaks. The primary concern of scanning in color is increased file sizes. Grayscale generally produces smaller file sizes and the use of compression technology can significantly reduce document size.
• Deskew: Automatically corrects crooked images in two ways using the document edge or analyzing the content of the images. The page straightening process is important for accurate conversion of your images. Commercial scanners deskew at the time of scanning. However images having a higher skew may require correction or a rescan.
• Noise Removal: Noise removal increases accuracy rates. The despeckle function cleans dots, specks and other noise from the images to improve character recognition. Despeckle is limited to bi-tonal (1-bit) images in such formats as tiff or fax.
• Image Enhancement: Image enhancement is used to improve poor quality images. It is best applied to smooth jagged edges and repair nicks on incomplete characters. Black and white (B&W) characters may also be thickened or thinned to improve recognition.
• Black Border Removal: Black border removal removes black edges surrounding scanned pages. It reduces processing time and improves the ability to zone text and pictures during batch recognition. Options included in border removal are border percent, white noise length and variance. The borders to remove may also me selected.
All documents created through OCR conversion are editable and have full text search capabilities. Organizations will find that it is generally very inexpensive compared to data entry service. The accuracy is usually near perfect for computer generated text-only documents and books. However, it may not be perfect for you and may not replace data entry service when the proofing required exceeds simple spell checking. This is usually the case for poor quality originals, poorly delimited spreadsheet data, lease documents with fine print, complicated layouts or books with pictures and graphics.
The only way to know if OCR will work for you is to test it with a sample. Then recommendations can be made to optimize the accuracy and improve the output. In cases with the final output is unacceptable a cost comparison to manual entry should be made.
Source:http://ezinearticles.com/?OCR-Spell-Checking---How-to-Improve-Object-Character-Recognition-Conversion-Results-to-Word-Format&id=4345862
• Scanning Resolution: Object Character Recognition works the best with documents scanned at 300dpi resolution. Scanning at higher resolutions and color scanning significantly increase the time it takes to scan a document. Increasing the resolution to 400dpi for spreadsheets, ledgers and old newspapers may improve the results. Typically however, you do not need to scan at 600dpi unless the pages have very small font sizes less than 6pt.
• Color Scanning: Deteriorated documents may also be completely unreadable when scanned in black and white (B&W) mode. Color and grayscale scanning improves the recognition rates for old documents that are yellowed, stained, wrinkled, and faded. Color capture may also improve the recognition rates of documents with color backgrounds, shading or small fonts and line breaks. The primary concern of scanning in color is increased file sizes. Grayscale generally produces smaller file sizes and the use of compression technology can significantly reduce document size.
• Deskew: Automatically corrects crooked images in two ways using the document edge or analyzing the content of the images. The page straightening process is important for accurate conversion of your images. Commercial scanners deskew at the time of scanning. However images having a higher skew may require correction or a rescan.
• Noise Removal: Noise removal increases accuracy rates. The despeckle function cleans dots, specks and other noise from the images to improve character recognition. Despeckle is limited to bi-tonal (1-bit) images in such formats as tiff or fax.
• Image Enhancement: Image enhancement is used to improve poor quality images. It is best applied to smooth jagged edges and repair nicks on incomplete characters. Black and white (B&W) characters may also be thickened or thinned to improve recognition.
• Black Border Removal: Black border removal removes black edges surrounding scanned pages. It reduces processing time and improves the ability to zone text and pictures during batch recognition. Options included in border removal are border percent, white noise length and variance. The borders to remove may also me selected.
All documents created through OCR conversion are editable and have full text search capabilities. Organizations will find that it is generally very inexpensive compared to data entry service. The accuracy is usually near perfect for computer generated text-only documents and books. However, it may not be perfect for you and may not replace data entry service when the proofing required exceeds simple spell checking. This is usually the case for poor quality originals, poorly delimited spreadsheet data, lease documents with fine print, complicated layouts or books with pictures and graphics.
The only way to know if OCR will work for you is to test it with a sample. Then recommendations can be made to optimize the accuracy and improve the output. In cases with the final output is unacceptable a cost comparison to manual entry should be made.
Source:http://ezinearticles.com/?OCR-Spell-Checking---How-to-Improve-Object-Character-Recognition-Conversion-Results-to-Word-Format&id=4345862
No comments:
Post a Comment