How to determine the correct coordinates of the block to recognize text in the picture?

Good day to all.

There is such task. This image of the scanned text. For the image coordinates, some areas that need to be recognized. These fields may not be accurate.

Is there any algorithm which allows to find a correct location of each block? I.e. use not the exact coordinates you specified earlier, and apply some "tolerance"?

OCR uses the Tesseract OCR.
September 18th 19 at 11:39
1 answer
September 18th 19 at 11:41
Solution
"ABBYY FlexiLayout Studio
ABBYY FlexiLayout Studio is a powerful tool to create flexible descriptions for documents with non-rigid structure, that is, those documents on which the location of the same data may vary from one instance to another. For flexible descriptions in the interface of the program it is called FlexiLayout. Flexible description FlexiLayout allows the system to extract from documents with non-rigid structure of the required data, such as text strings, barcodes, date, currency symbols, numbers, separators, tables.

For example, using only one flexible description of the program can find all the required fields in the accounts received from different vendors and having different format, even if it is a multipage document. ABBYY FlexiLayout Studio offers effective and convenient means for testing and setting flexible descriptions on the set of document images: a tree of hypotheses, reference descriptions, FlexiLayout language.

Implemented in ABBYY FlexiCapture technology-learning systems not only to automate the process of creating descriptions of the documents, and allow experts to edit the descriptions after creating them to improve the result of the operation of the system in the future. This greatly reduces the complexity of this phase of work and its duration."
Can change is a recognition tool?
In General, it is about what you need. But not quite. The point is that I have not finished the app, and an SDK or library that you can use in your project. Well, the description of the algorithms that allow to write the required functionality. - lurline_Shields commented on September 18th 19 at 11:44
Herself will respond. In addition they have a SDK. - lurline_Shields commented on September 18th 19 at 11:47
: I don't even know what SDK is, I'm the admin... Already know.
But that Feinrider have long forms recognize - know. - Camron_To commented on September 18th 19 at 11:50
you the advertising agent? - alfonso56 commented on September 18th 19 at 11:53

Find more questions by tags Text recognitionTesseractpattern recognition