The “pixel-by-pixel” strategy reads data from each individual black cell one-by-one, using a separate read operation for each required cell.
The “bounding-box” strategy (left) reads all the black and grey cells in a single read operation; the grey cells are then discarded. The “scanline” strategy (right) uses a separate read operation for all cells with the same j coordinate, on the assumption that each operation is reading a contiguous block of data from disk; again the grey cells are discarded.