Confidence Values
The confidence value of a result represents the level of certainty that MICR Xpress has in the correctness of the Text property of the result. A greater confidence value means that MICR Xpress is confident in the correctness of the Text property it has provided for the result. A lesser confidence value means that MICR Xpress is less certain about the correctness of the Text property it has provided for the result.
Confidence values are calculated on several factors depending on recognition characteristics. They should only be considered as relative and only to other results of the same type. For example, confidence values of a text line should not be compared on the same scale as confidence values of a character result.
The Confidence property of the result object provides the confidence value calculated by the recognition engine from the analysis of the MICR character, text line, text block, or MICR field. The value range is 0-100 where 100 is the greatest confidence.
How Confidence Value is Determined
When an image is analyzed by MICR Xpress, the MICR Xpress engine finds the best matching character class for each character in the image, using multiple neural networks. The confidence of an individual character result is computed based on the scores that each neural network gives for the similarity of the image to the character class. The confidence is computed in the same manner for any alternative results for a character result.
Confidence of a text line and text block is computed based on the aggregate confidence of all characters in the text line or text block. Confidence of a MICR field is also computed based on the aggregate confidence of all character in the MICR field. The confidence of the MICR field does NOT represent confidence in the correctness of parsing.
Increase Confidence Values through Methods and Properties
 |
Set and modify the properties tabulated below prior to using AnalyzeField. |
| Property | Description |
| MinimumCharacterConfidence | The value used for thresholding character results based on confidence. Empirically adjust this value as needed to obtain the required level of accuracy. In the case that the confidence result does not exceed the MinimumCharacterConfidence, the text of the character result is substituted with the RejectionCharacter and the result's confidence value is changed to 0. |
| RejectonCharacter | The character used as a replacement value when the recognition confidence is below the MinimumCharacterConfidence. When this character is used, the confidence of the result will always be 0. |
The confidence values returned for characters, lines, and blocks of text recognized by MICR Xpress are significantly different from those returned by SmartZone and SmartScan Xpress ICR/OCR/OMR/MICR.
If you have previously used SmartScan Xpress, it is important to re-evaluate your test images and re-calibrate appropriate confidence thresholds that your program may use in choosing various logic paths. Examples may include decision thresholds to apply or reject various image cleanup operations, or to initiate operator-assisted field validation.
If you have previously used SmartZone it is important to understand that the recognition engine and input data is different for MICR Xpress and therefore optimal settings for SmartZone’s OCR classifier may not translate into optimal settings for MICR Xpress.