A macroblock is the square pixel region a codec processes as a single coding unit - classically 16x16 pixels in H.264/AVC. It is the cell of the grid that becomes visible as blocking, and it is the origin of the synonym macroblocking. The macroblock is the large unit used for motion prediction; inside it the DCT runs on smaller transform blocks (4x4 and 8x8 in H.264), so the visible seams can fall on either grid. Newer codecs generalize the idea with larger, variable-size units: HEVC uses Coding Tree Units up to 64x64, and AV1 uses superblocks up to 128x128, each recursively split into smaller partitions. Because the modern grid is variable rather than a fixed 16x16, the practical, codec-neutral term is simply block - which is why this guide keeps the generic framing rather than calling everything a macroblock. The unit matters for measurement because no-reference blockiness detectors look for energy at the block-grid frequency, the inverse of the block size.
