The new capabilities of Noisemaster have been implemented in XCaliper 6.0 via the addition of two new tools (RPT-TeachPattern and RPT-TeachFont).  These classifier-creation tools duplicate all of the properties and methods of the existing NFT-based TeachPattern and TeachFont tools thereby allowing for an easy transition between technologies. In addition, the RPT-based tools include an invert variant method,  which allows users to create pattern variants based on inverted grey-scale values from the original patterns.

The Search tool has been modified to support the use of classifiers generated from any of the teaching tools. Users select the classifier source from the Search tool property pages by specifying the file type in the classifier-loading dialog. The two file extensions supported are (*.ncl) for NFT classifiers and (*.nmc) for RPT classifiers.

• Define the pattern. You define a pattern of pixel intensities in an image by giving the pattern a label and pointing out one or more examples of it in one or more images. The set of examples is called a training set.
• Teach the classifier. You create the classifier, a file containing all the information required to identify a pattern. The classifier is saved in a file or given a name, and used as a memory classifier in the Search tool. The classifiers are very robust, even those that have only been given one example of the pattern. Using the interactive property pages, you can create more complex training sets without writing a line of code.
• Test the classifier. The TeachPattern tool allows you to determine the accuracy of the classifier by testing it on new images.

These tools, TeachFont and RPTTeachFont, are specifically suited for creating OCR applications that read strings of characters.  

The TeachFont tools can also generate rotated and scaled variants of the characters you teach and include them in the training set used to create the classifier. Then the classifier can recognize the characters at different orientations and sizes.

These tools, like the TeachPattern tools, do not analyze an image and report results. The purpose of the TeachFont tool is to create a classifier the Search tool can use to recognize a string of image patterns, that is, many individual image patterns and the order in which they occur.

• Flexibility in defining the search area and direction. The viewport is a parallelogram defined by three points which specify the portion of the image to search, the general direction of the search, and the direction of the scan lines. You control these points by designating pixel coordinates or dragging the pick points of the tool on the template image.
• Rotation and scale invariance. The Search tool can transform the classifier by instructing it to find a pattern at a different angle, different size, or both.
• Use search results to set the coordinate transform. The search tool reports the position, angle and scale at which the best match was found. Your application can use this information to set the coordinate transform to match differences in translation, rotation, and scaling.
• Use a variety of search methods. The search tool allows you to select from a number of powerful searching methods. Search Optimum finds the single best example of a pattern. Search First finds the first example of the pattern above a quality threshold. Search All finds all pattern examples above a quality threshold.
• Multiple modes search. The search tool can simultaneously search for more than one pattern.
• OCR and OCV. The search tool can use classifiers that contain fonts. The ReadToken search method reads a string of characters. The ReadVerify Token returns a quality value for each character that is read. ReadExtent reads all characters within the search area.