This strategy is the inverse of naked pairs and can be used to clarify remaining candidates in the grid, leading to other solving opportunities. It applies to rows, columns and subgrids but is easier to explain by discussing one of those at a time.
You're looking for a pair of candidate numbers used only in two different cells within the same row. They are harder to spot than naked doubles as other candidates can also be present. As the pair of cells must contain both occurences of the paired digits (but we don't know which way round just yet), we can eliminate any other candidates in the paired cells.
N.B. You should only use this strategy once all candidates in the row (or column or subgrid) have been filled with candidates.
In the example, the numbers 4 and 5 form a hidden pair as they are only used in two cells of the row. We can then eliminate any other candidates in those two cells turning them into a naked pair.