Refactor code of OCR and VerticalOCR to utilise only two points in bounding boxes #760
Description
For now, they uses all four vertices of bounding box, but two (e.g. bottom left and top right are completely sufficient). The scope of this issue is to change implementation to use only two vertices, see comment here: #754 (comment)