April 2001

SCAN STATION

Process and Automate Color Images

by Maria Medina

Now that color scanning has become more affordable, perhaps you're thinking of switching to color imaging. The Prizm Color Image Processing Toolkit, from TMS Sequoia, Stillwater, OK, offers a battery of functions that help you use color information to automate document identification and data extraction.

Color is a mixed blessing for businesses that cope with paper-based transactions. Many organizations use color to draw customer interest and make documents more readable, but when it comes to processing, color can make life difficult. The reader response cards bound into so many magazines are a good example of this challenge. The colorful, compelling designs of marketing departments and art directors can make bitonal scanning and data extraction tricky. Color images are much more readable and forgiving, but until recently, few processing tools were available.

Prizm is a C++ toolkit that lets you integrate color image processing features into customized applications. The toolkit has five modules: the core module plus ColorDrop, ColorFind, ColorEnhance and ColorForms. The core module handles file input, thresholding and release (to another application), and it can be teamed with any of the other image processing modules.

The ColorDrop module lets you drop out one or more colors and deliver grayscale images to an image archive or to another application. There are two methods for color dropout. Virtual bulb dropout works in much the same way color bulbs are used to drop color information out of bitonal scans, but the approach is more flexible. Virtual bulb looks for peaks in a spectrum, so it can adapt to variations in color hues and extract the one you want, or dropout the selected color altogether.

The second method, proximity drop, lets you specify and block an area of an image from further processing. This feature can be used to protect vital information. For example, say you have a white and red form that was filled out in red ink. By selecting the red hue of the pen ink, you won't lose the data. Because the ColorDrop module can select precise hues, it can detect which variations of color it should not be processing.

Prizm's ColorFind module supports automated indexing using color. This works similarly to other color-enabled indexing tools, such as Colindex, from Dunord, Montreal, and Color ID, from Kofile, Rochester, NY (see "Color Makes Indexing Easier," July 2000). These tools save time by automating document identification and data extraction with the help of preprinted colors or highlighter marks. To automate document ID, you can key on preprinted colors or use different color highlighters during document preparation to send images into different workflows. To automate data extraction, you can highlight different data types with different colors - even on unstructured forms on which data appears in different locations. ColorFind automatically drops out the color highlight information and extracts and exports recognized data to another application. The rest of the image can be archived, deleted or delivered to any other application.

The ColorEnhance module cleans up color images by removing shadings or borders, deskewing, datacropping and extracting features. It also comes with "Image Detergent," which eliminates variations of color in your images and sticks with the most common shades. This way, according to TMS Sequoia, you can reduce a file's size but still have an image that closely resembles the original document.

The Prizm ColorForms module processes color forms for identification, registration, dropout and mark sense. For example, you can train the module to recognize certain forms by building profiles. On a reader response card, for instance, the module can determine the document type and apply the image processing techniques stored in the profile.

Corporate pricing for the core module paired with the ColorForms module is $60,000. The core module paired with ColorFind, ColorDrop or ColorEnhance costs $85,000. You can purchase the complete set of modules for $135,000. TMS Sequoia is also developing an ActiveX control for the toolkit.

Even if the information you need is delivered in bitonal text, color images offer more than just better image quality. The Prizm toolkit helps you deal with the challenges of real-world color documents while also giving you time-saving and money-saving features for automated document identification and data capture. TMS Sequoia is offering the right tools for color image processing, and with the addition of an Active X control and stand-alone applications based on the toolkit, it will soon reach a wider audience than the custom development market.