I use the pdfocr program with tesseract when I want to ocr my pdf's on linux
I use the ppa:gezakovacs/pdfocr repository for pdfocr and
sudo apt-get update
sudo apt-get install pdfocr
sudo apt-get install tesseract-ocr
sudo apt-get install tesseract-ocr-eng
The command to convert is
pdfocr -i input.pdf -o output.pdf
In addition, prior to converting I have to do
in order to tell tesseract where the language files are.
After conversion you can grab your text by copying it from the output.pdf itself, and from there to your LaTeX file.
Tesseract does have a math recognition language file which is listed as one of the language packs: equ(Math/equation detection module)
The libreoffice method only works in a limited way because it transforms pdf's into libreoffice graphics files (.odg) and this prevent you from copying text in big blocks, as text is ocr'd as short graphical blocks, not as a whole. It therefore needs to be copied sentence by sentence if at all, which is very tedious for more than a few sentences.
I have been looking for a pdf to TeX converter, and so far have not found one.