This is essentially the inverse of the pointing technique and again allows certain candidates to be ruled out and can be applied to both columns and rows.
For a given row, if all occurrences of a candidate digit on that row are also within one subgrid, then the candidate cannot be used elsewhere in the subgrid. The row essentially "claims" the candidate within the subgrid. The same is true for claiming columns.
N.B. You should only use this technique once all of the cells on the row (or column) have been filled with candidates. If there are gaps then the candidate digit in question might appear in those gaps which can lead to errors and the puzzle won't be solvable!
In the example, candidate digit 4 appears on the last row only within the bottom-left subgrid. This means 4 cannot appear elsewhere in that subgrid.