In order to prohibit OCR of PDFs which already contain a text layer, please use
ObjectsExtractionParams.SourceContentReuseMode = CRM_ContentOnly
Please note, that even in this case, characters that were not present in the text layer of the source PDF document, may appear. This happens if for any text on the image layer there is no corresponding text on the text layer. FREngine recognizes these parts of the text.