01.   What is Purging?

Part of the process of converting paper to images requires an evaluation of the material involved. Some file folders, documents, or other repositories contain extraneous material, duplicates, notes, and other information that need not be scanned. In these instances, you must decide whether it is more cost effective to purge files before scanning, or to scan everything and purge extraneous images. In some instances, purging requires the advice of personnel with a knowledge of the documents being scanned; e.g., purging drafts or other old versions of documents. We call this subjective purging. In other instances, purging can be done by persons without such knowledge; e.g., purging all handwritten notes, purging all notes, Post-It notes, etc. We call this objective purging.

We do not recommend subjective paper purging; instead, we recommend that subject purging be done with images--subjective image purging. Subjective image purging means that a cost was incurred in scanning the images; however, that cost makes the purge process much more efficient and inexpensive. In other words, scanning and throwing away images is less expense that purging paper files.

We recommend objective paper purging. Since objective purging specifies exactly what is to be throw away, it can be done quickly and easily when paper is being prepared for scanning.

02.  How does Organizing help in scanning?

Like pieces of paper, images typically are grouped into documents. Accordingly, the beginning and end of each paper document must be clearly defined to maintain its integrity after conversion. We do this with document separator pages.
Document separator pages are inserted between documents during the document preparation phase. They typically have a bar code or "patch code" printed on them. This code tells our software tells our software where one document ends and another begins.

In some cases, we encode indexing information in the bar code on the separator page. This can be a file number, a client name, date, etc.. When job constraints allow us to use this technique, we can read the bar code and use the information to automatically populate the index fields associated with that document without human intervention. This can greatly reduce the cost of indexing documents.

03.  What do you mean by Paper Preparation?

A certain amount of physical preparation is required to prepare paper to be scanned. For instance, if a job is to be undertaken using high-speed autofeeders, binders like staples, brads, and paperclips must be removed, as should Post-Iit notes and other attachments. Depending on the job, it may be necessary to rebind documents after scanning.
Physical preparation also typically requires that paper be "jogged" so that all leading edges are aligned and ready to be fed into the scanner. These simple but necessary steps help eliminate scanner jams and double-feeds.

Paper size and weight also should be considered. Many scanner auto-feeders cannot handle mixed widths and weights and have specific width and weight limitations even when the paper is uniform. Accordingly, an appropriate scanner and feeder must be selected to match the condition of the purged, organized and physically prepared paper. In some cases, flatbed scanners may be required.

Finally, consideration should be given to "batching" documents. Batching helps to both control and improve the efficiency of the conversion process. Batching provides a convenient way to audit the process by matching scanned batches of paper with corresponding batches of images. Batching also can be used for other quality assurance checks, such as batch scan count comparisons (paper batch count compared to image batch count), batch tracking through the conversion process and batch log files with all information on images captured and indexed during the conversion process.

04.  Image Quality

Document scanners typically produce a black and white (bitonal) image. Grayscale scanning while technologically simple, generally is not employed because file sizes are orders of magnitude larger than bitonal file sizes.

The initial question on document scanning is whether a bitonal rendering of your paper will be satisfactory. The answer is a simple "yes" where the paper is white and the text, black. However, more careful consideration must be given to documents like invoices, where gray backgrounds or red or green colored boxes are common; photos, where bitonal renditions offer significantly less detail or documents containing photos and text.

Most image quality issues depend on the scanner selected to do the job. Some scanners render black text on white paper flawlessly, but do a very poor job where colors or grays are part of the requirement. Other scanners handle very difficult gray and color requirements nicely using a process called dynamic thresholding. Still other scanners allow two or more images to be captured from a single piece of paper, with one image capturing the whole page at a low resolution, and another image capturing just a portion of the page at much higher resolution.

The choice of the scanner is critical to the issue of efficiently producing high quality images of the particular paper to be scanned.

05.  What do you mean by Paper Handling?

Paper handling also must be considered. Different scanners use different paper transports. Some use belts, others ball-bearings or use rollers. Auto-feeders also use various paper handling techniques and have different limitations. How a scanner handles paper has a direct impact on its suitability for a particular project. For instance, many scanners can not handle onion skin, card stock or batches containing mixed paper widths and weights. Others can't handle small paper or paper wider than 8.5 inches.

Some scanners support manual feeding in a way that is much faster than others. This is an important consideration if you know that the paper to be scanned cannot be handled by auto-feeders.

Like image quality, the key to fast, efficient paper handling depends on selection of the right scanner.

06.  Explain Image Resolution

Image resolution determines the number of pixels, or dots, per linear inch. Popular resolutions include 200 dpi, 240 dpi, and 300 dpi, though 400, 600 and even 1200 dpi.

A 300 dpi resolution renders 300 dots per inch. Higher resolutions generally improve image readability, though there are issues that must be taken into consideration. For instance, many monitors display images at 72 dpi, regardless of the resolution at which they were scanned. Even high-resolution monitors display at only around 200 dpi. In both instances, you must "zoom in" on the image to view it at the resolution actually available. Similar considerations relate to printing. A 200 dpi image will print the same as a high-resolution 600 dpi image on a printer that only prints at 200 dpi. Scale-to-gray technology makes this subject even more confusing. It employs special techniques to render images more readable by using techniques like dithering. Dithering extrapolates from available resolution information to make images much more readable with less jagged lines and more complete characters.

Image resolution also significantly effects image file sizes. For instance, the file size of a compressed TIFF Group 4 compressed image at 200 dpi might be 60 KB, while at 300 dpi, it might be 90 KB.

We make our resolution recommendations based on the level of detail required to capture all needed substantive data from the paper, while minimizing file sizes.

07.  What is Image Deskew?

Two skew issues are involved in document imaging; paper skew and print skew.

Paper skew relates to the relationship between the paper and the scanner camera as the paper is scanned. If the paper is skewed, the image is skewed. Paper skew typically is introduced with scanner auto-feeders and to a lesser extent, their internal paper transports. Some scanners control paper skew well. Others do not.

Print skew relates to how the print actually was deposited on the paper. Print skew relates to the relationship between the print and the paper on which it is printed. Photocopied and faxed documents tend to have skewed print.
Paper and print skew effect document images in two ways. Skewed image text is less legible and is not processed well by OCR engines.

The solution is to electronically deskew the image by re-orienting image pixels along a corrected x/y axis. This technology usually is very effective; though in a small number of cases, it can introduce unacceptable distortion.

08.  What is Image Border Cropping?

Depending on the scanner and scanner control software employed, it may not be possible to exactly and automatically match the size of the captured image to that of the paper scanned. For instance, a software solution that requires you to manually define the image size to match the size of the 8.5 by 11 inch paper you expect to scan will capture an 8.5 by 11 image even if some 4 by 6 inch cards are mixed in. In these cases, an ugly black area will surround images.

Image cropping removes extraneous black borders--either by requiring a human operator to manually define the area to be cropped, or by employing sophisticated algorithms to evaluate the image and automatically crop borders.

09.  What is Noise Removal?

Scanners often interpret minor paper imperfections or extraneous dots on paper as small groups of black pixels called background noise. Carbon forms are excellent examples of paper with significant amounts of background noise.

Background noise makes bi-tonal images less legible and image file compression schemes much less efficient.

Noise removal algorithms examine an image, identify likely black pixels constituting background noise and convert them to white pixels. The result is a much more legible image and a much smaller compressed file size.

10.  What is Background Removal?

Documents can contain vertical lines, horizontal lines and background shading that represent no substantive data. In these cases, it can be desirable to remove them, since doing so can make the image more legible and dramatically reduce file size.

Background removal algorithms are available for this purpose. Care must be taken in using them, however, since it is not always easy to predict when a vertical or horizontal line might in fact be critical in conveying the data represented in an image. We recommend the technology only where all images to be processed using background removal techniques have been tested and the results evaluated.

11.  How does Annapolis Technologies implement Process Quality Control?

Whenever possible within a conversion process, Annapolis Technologies uses technology to replace typically labor intensive quality control processes.

When bar code separation sheets are used in processing documents, we include a quality control process to verify that each bar code has been read by the system. Since we produced the bar codes, we know which bar codes we should find in each batch. Even the best bar code readers on the market will miss a small percentage of bar codes just as the bar code reader at the grocery store will miss some. The Annapolis Technologies quality process is to compare the list of expected bar codes with the list of captured bar codes. Human intervention is only required if the two lists do not match.

Annapolis Technologies uses a process called NIC-VIC (Number Image Count - Verify Image Count) to ensure that each and every page given to us for scanning is scanned. The patent-pending process is simple yet powerful.
Annapolis Technologies uses the best production scanners and counters on the market. Using High quality equipment to do the work at production speeds, Annapolis Technologies enables us to do better quality work at more competitive prices.

The key to effective image-enabled data entry is setting the job up properly and employing all appropriate data extraction and validation techniques.

12.  Explain Full Text OCR/ICR Processing

Images are useful only if they can be found when needed. There are four common ways to address this issue:

• "Filing" related images in subdirectories or electronic folders.
• Matching images with index fields or keywords in a structured database
• Linking images using hypertext links.
• Matching images with text files in a full-text database.
  

If images are to be found by searching a full-text database, a machine readable ASCII text version of the image must be created. This is done using OCR (optical character recognition) or ICR (intelligent character recognition) processing engines. These engines are useful for full-text processing only if the text is machine print. They will not produce acceptable results from hand printed or cursive data.

The text produced by these engines from machine printed data can range from extremely accurate to very poor, depending on image quality, resolution, type faces and the OCR or ICR engine employed. We recommend a careful analysis of your documents before making a decision on whether to full-text OCR process them.

13.  Explain Form OCR/ICR Processing

Form OCR/ICR processing differs from full-text OCR/ICR processing in that it does not attempt to translate an entire image into ASCII text. Instead, form OCR/ICR processing attempts only to translate image form fields located at specific defined image coordinates. These field coordinates are always located at the same places on a given form. Since fields are part of a form, steps can be taken to control the way data inside the coordinates is presented. For example, the form can prompt users to print, to print within boxes and to print only in black or blue ink. Handprint recognition is quite practical under these kinds of constraints.

Since form OCR/ICR processing deals with known data fields, it can incorporate a number of techniques to improve and ensure accuracy even from poor quality images or difficult handprint. For instance, data extracted by OCR/ICR engines from known data fields can be compared to tables defining acceptable types of data; alpha, numeric, date, zip code, etc. to help the OCR/ICR engine interpret the image pattern. Similarly, post-OCR/ICR routines can be used to correct characters flagged as questionable by the OCR/ICR engine. Finally, human operators using a variety of display options can quickly and efficiently review and correct OCR/ICR results.

Form OCR/ICR processing by itself or assisted by human data entry editors and verifiers can reduce key entry chores by orders of magnitude.

14.  What is Form Dropout?

One of the difficulties in using form OCR/ICR processing relates to the manner in which people fill out forms. They often ignore form instructions by printing "outside the lines" or over portions of the form itself. Data "outside the lines" will not be within the OCR/ICR zone. If the zone were enlarged to include it, interpreting the data still would be difficult because form lines and instructions would degrade it.

Form dropout techniques address this problem by removing the form.

This is accomplished by matching each image scanned to a library of blank dropout form patterns that have been scanned and stored for comparison. When an image contains a dropout form pattern, the form dropout algorithm removes it, leaving only the data that was entered onto the form. At this point, zone OCR/ICR techniques can be used to translate the data in the identified coordinates into ASCII text.

Form dropout is useful for two reasons:

• Like other removal techniques, it dramatically reduces file sizes
• OCR or ICR engines can much more efficiently and accurately analyze and convert an image into ASCII text with the form removed.
  
  

The actual images in a form dropout scenario can be managed in three different ways:

• The original image can be preserved, while the dropout image is deleted after its data is extracted
• The original image can be deleted, while the dropout image is preserved in a manner that combines it with a single overlay image of the blank form when the dropout image is retrieved and displayed
• The original and dropout image can be deleted, while the ASCII data extracted from the image is stored and presented as either ASCII data or as ASCII data with a single overlay image of the blank form when the data is retrieved and displayed.

15.  What is the standard for image Files?

As a result of the trend towards standardization in the imaging industry, the overwhelming image standard is TIFF G4 (tagged image file format; TSS Group 4 compression). However, there are some significant exceptions, most of which involve large imaging system architectures implemented in the early 1990's, large IBM imaging systems still being sold today (ImagePlus, Visual Info, etc.), very large image files from E-size drawings and JPG or GIF files used with color and grayscale images.

We recommend checking your existing infrastructure to verify that the TIFF format is appropriate. If it is not, the required format must be identified. Matching proprietary image headers is the common solution. This is not particularly difficult. Matching proprietary image compression schemes, though uncommon, can be extremely difficult without the cooperation of the original developer.

16.  How are text files used in the Document Conversion process?

Text files and variations of text files like RTF (rich text format files), store ASCII output from images that have been OCR processed. They also often are used as a temporary way to store indexing information relating to an image or batches of images until the information can be uploaded to a database.

We can present OCR output in text files, RTF files and other popular formats; however, the OCR engine employed to produce the files directly impacts the file format options available, as well as the speed at which OCR processing can be performed.

Text files containing indexing information also can be presented in a variety of formats employing carriage return, comma and semi-colon delimiters, specifically defined row/column formats, and more. The proper format choice will depend on the spreadsheet, database, or imaging system into which the data is to be uploaded.

17.  What are Objects?

Objects are files that contain not only data, but information about or components of the application software used to process the data. Many imaging systems "wrap" TIFF image files with code to turn them into objects. This allows the imaging system to more easily associate system-related data with the file.

Annapolis Technologies reformats images for object-oriented imaging systems when the images are uploaded into the system. This strategy has two advantages:

• It allows us to use TIFF images during the image and data capture process, which means we can take advantage of all of the industry-standard TIFF-oriented image processing and data extraction tools.
• It allows us to take advantage of the import tools designed and developed specifically for the object-oriented imaging system into which the images are to be loaded. Using native tools in this fashion insures image and text compatibility.