DpuScan « previous page | | next page »

goldgold Area-Dependent Dynamic Threshold Method

DpuScan is equipped with MDIT-functionality(Multi Dimensional Intelligent Thresholding). MDIT performs a three-step analytical process to determine the optimum brightness distribution in an image. This analytical process results in a crisp black & white image, no matter what the original looked like.


Here, MDIT works depending on zones and changes its strategy depending on the environment. While other solutions just stop the scan process for difficult images in order to allow manual corrections, MDIT never stops the scan process because it can make corrections automatically:


Can you see the big 5000 in the left gray image? If, on occasion, the results are not optimal, ReScan on Demand (RSoD) allows you to later create, from the gray image or even from the color image, an optimal bitonal black & white image on your screen - without the need to ever halt the scan process.

While other programs only offer setting possibilities for the scanner, DpuScan has its own threshold modules. You also have the option of using the functions of your scanner, however.

In order to generate a bitonal black/white image from a gray image, a decision is required for every gray value pixel whether it shall be converted to white or black. Easiest is to work with a fixed threshold value. Above this value, all pixels become white, and below it they turn black. Such a simple method is integrated to DpuScan as one possible method. The results, however, can be used only if good originals were scanned. In all other cases it will be better to constantly adopt the threshold to the environment conditions. DpuScan offers several strategies to define ACDP threshold filters, as follows:

The "Static Level" value determines the value below which a target pixel turns black anyway.

"Matrix Size" sets the number of pixels that the ACDP threshold algorithm inspects around the original pixel, in order to determine the mean background gray value. As the algorithm reacts only to differences in brightness, uniform dark areas are hollowed and represented just in a frame. Texts that were contained within this dark area, printed still darker, are thus conserved. An increased matrix can limit this effect.

"Dynamic weight" determines how strong the ACDP threshold algorithm shall react to local brightness fluctuations.

"Noise Removal" adds another algorithm to prevent the ACDP threshold algorithm from reacting to isolated pixels. It inspects the actual pixel and its surroundings. Use "Area size" to set the size of the surroundings to be inspected (normally, "Small" should be sufficient here). "Weight" influences the working intensity of this part.

When enabling the "Backtracker", the threshold values will automatically be adapted to fluctuating background brightness values. False parameters, set for ther static threshold, are automatically corredted. This way, most documents are automatically binarized correctly with one only setting, even at very high quality fluctuations.

With "Preview" enabled, you can observe the effect of filter setting immediately on the screen. Additionally, it is possible to enlarge the points of interest. Also the window size can be modified.
The above screen shot illustrates the threshold results for a gray image that was not pre-filtered by a color filter. Also the frames resulting from it can still be suppressed by pre-filtering with an ACDP color filter. Please take the results of ACDP color filtering and ACDP threshold, executed one after the other, from the following illustrations:

Threshold: Method 1 Threshold: Method 2 Threshold: Method 3

The above 3 examples are results you can gain from the same original, using the methods of our competitors. Method 1 is part of the Twain driver of a scanner manufacturer; it also increases the maximum scanner resolution from 150dpi to a calculated resolution of 200dpi. Methods 2 and 3 work directly on the color image.

All color images shown in the other illustrations were scanned with a Scamax 410S scanner at 200dpi resolution. In August 2003, the speed was 100 pages per minute (200 images at duplex scanning).

click into the images « previous page | | next page »

Janich & Klass · Computertechnik GmbH · Zum Alten Zollhaus 24 · D-42281 Wuppertal