Text Recognition (OCR)
AI OCR (file-based)
The AI OCR (file-based) Action Step is a further development of the existing OCR Action Steps. In contrast to the other OCR Action Steps, it is based on concepts of neural networks. The available languages, therefore, are based on trained patterns that take account of the characteristics of the relevant language. In many cases, therefore, the correct recognition of texts that are, for example, barely legible or have little contrast or an unusual font, is improved with this Action Step.
Using the AI OCR (file-based) Action Step, existing images such as document scans can be recognized.
The basis for this Action Step is an image file. This can be available in the regular graphic formats such as BMP, JPEG, PNG and TIFF.
You should consider the Intelligent OCR 2020 Action Step if more than one language is to be recognized at once or if you wish to use your own dictionary.
Properties
OCR Image Preprocessing
-
Force Black Background Reverses the color of the screenshots. This is especially recommended for white text on a black background.
-
Scaling factor Displays the selected enlargement in the wizard.
-
Remove Shadows If there are uneven shadows on the image, you can remove them with this option.
-
Noise Filter If there are disruptive pixels (image noise) in the image, you can suppress them with this option.
-
Enhance Contrast If an image has low contrast and is very pale as a result, you can use this option to increase the contrast.
-
Color Depth The digitalization of an image consists of two steps, rasterization and quantization. Rasterization refers to breaking down the image into small pieces. An original dot becomes a pixel. In the case of scanned templates, due to the low resolution of the scanner, not all original dots are converted. A reduction in resolution takes place. In quantization, it is determined which value is assigned to a pixel. If we only distinguish between black and white, we get a binary image. Each pixel is stored as one bit, we refer to 1-bit color depth. Typically, an image has a resolution (or color depth) of 8 bits. If there are several levels of brightness, we refer to grayscale values. Every other value that differs from 1 bit reduces the image to a lower grayscale value. This way, it is often possible to remove disruptive image artifacts, which simplifies processing with OCR. We recommend a value of 3 or 4.
Bypass processing pipeline With this option, you can completely deactivate image processing. The settings made in the wizard are not taken into consideration.
OCR Settings
Area mode Appears as soon as you define the search area under Scan Area. You can choose between:
-
Actual Pixels The image is used in its original resolution
-
Fit to DIN A4 The original image is enlarged in height or width until it corresponds to the aspect ratio of a DIN A4 sheet. Depending on whether height or width is adjusted, there are areas that are augmented. The original resolution of the image remains intact.
-
Fit to DIN A3 Behaves as with DIN A4. Since DIN A4 and A5 have an identical aspect ratio, there is currently no difference. However, in the future, new algorithms are likely to have to take account of the size. Therefore, this selection is now available.
Image file name* Name of the image file.
Image folder path* The path to the folder in which the image file is located.
Orientation If you selected Fit to A4 or A3 in Area mode, you can select the format.
-
Automatic
-
Landscape
-
Portrait
Inbound Variables
Image file name* Name of the image file.
Image folder path* The path to the folder in which the image file is located.
Outbound Variables
IsSucceeded (Boolean) Indicates whether text recognition has been carried out successfully or not.
Wizard
You can use OCR Image Preprocessing to improve the quality of the image so that the OCR operation can run optimally. This is always carried out in the specified sequence. Here, you can also switch the individual steps on or off.
If you click on Load design-time file, the Windows Explorer opens and you can select a sample file. You can start a test run with the Test button. For a local Workflow run or the subsequent productive process run, the Directory Path and file name must be entered correctly. The sample image selected for the test run is no longer considered in this case.
Under Directory path, you define the path to the new file to be created. You can create an absolute path or a path relative to the Base Path. You select an existing path To construct the path, you can also insert variables from the Workflow.
You can find further explanations under Properties.