Scaling Factor

The Scaling Factor parameter allows you to increase or decrease the number of pixels for IntelliText OCR to process in a given OCR character. This parameter is used to change the sampling interval of the ROI (region of interest).

 

IntelliText OCR is optimized for detecting characters in the range of 25-50 pixels in height. Optimal character width is 30 pixels. If your application requires the reading of text at high resolution, particularly high resolution dot print, this parameter can improve character detection by reducing character scale.

 

Because reducing the sampling introduces some loss of information, you may need to configure the optical setup accordingly. Scaling provides some speed enhancements when using larger ROIs and relatively low character counts.

 

If a symbol is very large or very small, adjust Scaling Factor to make the character height equal to 30 in height, which is the best height for the algorithm. Example: If the character is 60 pixels tall, set scaling factor to 0.5.

 

 

Scaling Factor Decrease – Example 1

The large dot mark in this example has excellent contrast and does not require optimization using binarization, but it does have segmentation problems. This particular mark is 130 pixels high, which causes the text segmentation not to identify entire characters properly. By reducing the scaling to 0.5, the character height is reduced to 65 pixels and proper segmentation becomes possible. Note, however, that a symbol such as this should be captured at a lower resolution for optimal performance.

 

 

Scaling Factor Decrease – Example 2

The following large mark shows image noise and a wide dynamic change across the ROI. The automatic binarization selection works well, but the low signal-to-noise ratio on the left side of the mark causes the C to fail. This could be optimized by fixing the binarization.

 

 

 

As this mark is approximately 80 pixels high, it is rather large for optimum text segmentation. Because the mark is so large, the algorithm has detected segments from the C on different lines of characters.

 

 

 

The segmentation problem is resolved by modifying scaling. This low-contrast mark may experience some instability from image to image with binarization set to Auto, and would likely benefit from a Fixed Template setting.

 

 

Scaling Factor Increase Example

In the following example, the dot mark has close characters and partially touching characters. The mark is at reasonable resolution but the close characters are creating segmentation problems. Character thresholds can be modified to separate them, but this can be challenging, particularly in the JUN segment. Setting the scaling to 2.0 and over-exposing the image creates a good separation that binarizes well and subsequently segments properly and consistently.