Paragraph/Block Reading Order of Text

[Mohit Jain]

I'd like to know what algorithm/heuristics Tesseract follows to determine the order in which blocks of text are read? Analysing the output of Tesseract on complex layout documents, I see that its not a simple row-order/column-order, rather some sort of hybrid fusion of the two. Can someone throw light on what happens under the hood?

Thanks,
Mohit