Prizm IP

Advanced Image Processing Toolkit

The Prizm IP toolkit offers the most comprehensive set of image processing development tools available for colour, greyscale and bitonal document images. Available as either an ActiveX/COM or C++ toolkit, Prizm IP’s powerful image cleanup technology is used to improve the quality and legibility of TIFF and JPEG images, reduce file size, improve recognition accuracy for forms processing applications, form identification, and form registration.

Prizm IP’s greyscale and colour image processing technology harnesses valuable information from scanned colour document images to deskew, remove borders, drop and extract multiple colours, use colour to sort and index and support greyscale and colour forms identification, registration, extraction, and much more.

Please Note: ScanFix Xpress v5 and FormFix v2 currently contain subsets of features offered within Prizm IP. ScanFix provides bitonal image processing. FormFix provides bitonal template-based forms identification, registration, extraction and mark sense. Pegasus Imaging highly recommends evaluating ScanFix and FormFix, unless you require Prizm IP for greyscale and/or colour image processing or form processing. The Prizm IP technology is actively being upgraded by Pegasus. The name Prizm IP will be phased out and the technology will become part of 4 products: ScanFix Xpress Bitonal v5 (available today), FormFix Bitonal v3 (which will be next), ScanFix Xpress Spectrum v5 (colour, greyscale and bitonal features), and FormFix Spectrum v3 (colour, greyscale and bitonal forms processing features).
 

Features

• Programming environments: Win32 visual development environments
• Sample code is included for: C#, VB.NET (Interop), VB6, VC6
• Can be used in any development environment that hosts ActiveX/COM controls or C++ libraries
• All features are available in the ActiveX/COM toolkit
• A subset of the ActiveX/COM toolkit features is currently available within a C++ library
• Process as many as 400-700 document images per minute (2.4 GHz processor, 512 MB RAM) when deskewing, cropping or performing noise or line removal

• Accurately deskew colour, bitonal and greyscale images
• Returned values will reflect the detected amount of rotation and the confidence value
• Prizm IP deskew does not cause laddering (common in other deskew methods), thus providing cleaner images
• Bitonal deskew with confidence values can process over 400 images a minute for 200 dpi images (running on a 2.4GHz processor with 512 MB RAM)
• Greyscale and colour image deskew speeds vary with file size, so DPI plays an important role in speed, but for 200 DPI images of business documents you can expect to deskew 130 images per minute in greyscale and 70 images per minute in colour
  
  

1. Accurately crop colour, bitonal and greyscale images
2. Data Detection with Confidence Values
   • Use when the background colour and paper colour in a greyscale or colour document image are the same
   • Use the resulting data detection to centre the data in the image, or set margins to a given value, or cut off excess border
   • Confidence values allow you to route exception images (images with low contrast or edges that are not sharp and straight) to another cropping method or to an automated operator correction station for exception handling
     
     
     
      
3. Rectangle Crop with Confidence Values
   • Find rectangular objects surrounded by a solid or nearly solid border, of a contrasting colour, in colour and greyscale images (i.e. a dark page on a light background, a light page on a dark background, or a blue page on a green background)
     
     
     
      
4. Border Detection with Confidence Values
   • Crop colour and greyscale documents of irregular shape
   • Find the edges of a scanned image by automatically comparing the object to a specific background value
   • Ideal for cropping microfilm/fiche document scans
     
     
     
      
5. Intelligent crop for bitonal images
   • Remove black borders
   • Remove white borders
   • Reduce image file size
   • Especially useful on microfilm or small documents that have been overscanned
     
     BITONAL IMAGE CROPPING
     
       
     
     BEFORE (file size: 100k)AFTER (file size: 15k)

1. Convert a colour or greyscale image to bitonal
2. Fast, accurate thresholding improves OCR read rates
3. Fixed thresholding - Allows the programmer to set the threshold value, all pixels above that value are converted to white, those below become black. Very fast, but can produce inferior results.
4. Dynamic Global - Analyses the image to find a threshold and then globally applies that threshold to all pixels in the image
5. Dynamic Local - Analyses the image to find a threshold, adapts the threshold for the local region around each pixel and then applies the individual thresholds to each pixel in the image.
   
   
   
   BEFOREAFTER
6. Invoice Thresholding
   • Clean up background gradients, regions of inverse text and noisy carbonless copies
   • Inverse text correction for colour regions improves OCR read rates of header data
   • Automatically adjust for background gradients and regions of inverse text
     
     
     BEFOREAFTER

1. Both automatic and manual brightness correction available
2. Global Brightness Correction
   • Designed to work with document images, automatically pushing the background towards white and making the data darker
3. Adaptive Brightness Correction
   • Designed to get more information out of photographic images, performing brightness mapping across the image
   • Designed for gray or colour images, to increase the contrast and brighten the image when the background has white scratches, banding or other variations
     
     

• Automatically rotate colour, greyscale and bitonal document images
• Identify likely text regions and pass those clipped areas to your OCR engine, determining rotation based on confidence values
• Images with low confidence values can be routed automatically for operator review
• Improve the speed of rotation by allowing Prizm IP to select small regions of likely text

• Blank Page Detection supports bitonal (black/white), greyscale and colour document images
• Automatically identify and remove blank pages
• Use blank pages as separators for multi-page TIFF documents
• Choose to ignore information in page margins (like hole punches), with settings that decide how "dirty" the blank image can be
• Use settings to ignore border data on over-scanned images and choose to identify pages with minimal content as blank

1. Drop and extract multiple colours in a single pass
2. The Colour Drop module offers two methods for dropping and extracting colours, supporting greyscale as well as full colour images: VirtualBulb Drop and Proximity Drop.
3. VirtualBulb Drop - drop and extract specific hues from a colour image to mimic colour drop bulbs in scanners, returning a greyscale image
   • Powerful function for forms with white backgrounds
   • Retain the original image for archive or forms analysis reasons, while also being able to produce the best possible image for OCR
   • Drop or extract wide or narrow ranges of colour
   • Choose multiple hues to be dropped simultaneously
   • Adjust for variances in the hue to be dropped, allowing you to adjust automatically for scanner changes over time.
   • The output is a gray image
     
     
     
      
4. Proximity Drop - drop or extract specific RGB colours from an image, leaving other, similar shades in place. It affects only the colours closest to a selected colour in the RGB colour space.
   • In an image where bright, saturated red needs to be dropped, but pink and dark red need to remain, Proximity Drop would convert only the bright, saturated red. The pink and dark red will remain. Proximity Drop extracts colours, and also drops colours with no hue, such as black, white, and gray shades. Proximity Drop is the best choice when working with forms that have a background other than white. Shown here is the removal of a specific shade of blue. Notice how the second box is now white. Proximity Drop gives you more control, but allows less room for error in choosing the actual colour range to drop.
     
     
      

• Sort and index using colour
• Report colour values and locations to improve automated processes
• Locate regions of colour on a page, allowing you to use the information to sort or index document images based on colour content, regardless of whether you plan to archive the document image in colour or in black and white
• TMSSequoia research has shown that coloured markers can be a very reliable way to mark black and white documents for sorting. You can also locate highlighter marks, extract the highlighted text and drop the highlighter before thresholding for archival purposes.

• Specifically designed to improve colour document images for viewing or image processing
• Reduce background noise while preserving characters
• Sharpen images by finding the most common colours and removing the variations in those colours
• Can also be used with specific colour values, and is often useful when dropping paper colour before thresholding an image
• Can remove the variations caused by image bleed through without degrading individual characters
• Can decrease file size up to 20% depending on the image and scanner used, without impacting OCR results like many smoothing filters
  
  

• Reduce noise on colour or greyscale document images
• Filtering allows you to change the values of individual pixels based on the values of surrounding pixels. There are several predefined filter types, plus you can define your own filters to apply to the image.
• On colour images, these filters work on the red, green, and blue planes individually. Select a filter type in the table below for an explanation and example of a processed image.

1. Decompression (viewing):
   • TIFF, JPEG, JPEG 2000, CALS / C4 / JEDMICS, GIF, PCX, BMP, and more
2. Compression (and conversion):
   • Bitonal: Group 3, Group 4, LZW-Uncompressed
   • Colour: JPEG, LZW Packbits, Raw

1. Support for Greyscale, Colour and Bitonal image processing
2. Identify thousands of forms with accuracy
3. Manage thousands of templates and add new templates on the fly
4. Template-based form identification, registration, colour mark sense and Intelligent Document Language (IDL).
5. Add on SmartScan Xpress Barcode and/or SmartScan Xpress ICR/OCR/OMR to develop extremely accurate and powerful forms processing applications
6. Structured Forms
   • Used to create, design and publish your own paper forms
   • Used when you know where all relevant fields are and have a certain amount of control over format.
   • Design forms for your scanner and software (i.e. retail lockbox coupons, applications, tax forms).
   • Includes extremely accurate forms identification and registration, and powerful extraction methods that adjust for small variances in form data generated by the printing or scanning process
   • So powerful, the underlying technology was used to register forms in the largest data capture project in history (the Decennial Census).
7. Semi-Structured Forms
   • Used when paper forms originate from a variety of sources, often in different sizes and formats, but mostly containing the same information (i.e. purchase orders, invoices, medical claims (HCFAs, UBs), explanations of benefit (EOB), explanation of payment (EOP), shipping documents, bills of lading, customs declarations, etc.).
   • Semi-structured forms assume that the data will be relatively consistent from form to form but the location of the data will vary by form originator.
   • Identify a form from a table containing thousands of forms in a few seconds.
   • New templates can be added "on-the-fly," while old templates can be aged out of the system.
   • Based on keywords and fields, identify the most likely regions for data to be extracted and present the template to an operator for verification if necessary.

• Identify company logos or other features on a document image
• Use as a recognition aid with template identification when more than one company uses the same invoice template
• Use to help locate features on a page for quick form set-up
• Reliably find bubbles on test forms and boxes on questionnaires

• Quickly identify regularly shaped objects like timing marks on forms for fast ID and registration purposes.
• Use projections on printed forms that contain timing marks to increase the throughput of your forms processing system, while having the backup of Prizm IP's full page identification for images which return low confidence values.

• Identify regions of image content and return their locations
• Process data by zones or by characteristic
• Intelligently extract even the most difficult data

For more information please contact the MicroWay sales team:
Head Office MicroWay Pty Ltd PO Box 84, Braeside, Victoria, 3195, Australia Ph: 1300 553 313 Fax: 1300 132 709 email: sales@microway.com.au ABN: 56 129 024 825	Sydney Sales Office MicroWay Pty Ltd PO Box 1733, Crows Nest, NSW 1585, Australia Tel: 1300 553 313 Fax: 1300 132 709 email: sales@microway.com.au ABN: 56 129 024 825	New Zealand Sales Office MicroWay Pty Ltd (NZ) PO Box 912026 Victoria Street West Auckland 1142, New Zealand Tel: 0800 450 168 email: sales@microway.co.nz
International: call +61 3 9580 1333, fax +61 3 9580 8995