Quadtree partitioning is the technique HEVC and similar codecs use to recursively split larger blocks into smaller ones. Start with a 64×64 CTU. The encoder asks: "is this CTU best coded as one big block, or should I split it into four 32×32 sub-blocks?" If split, each 32×32 sub-block recursively faces the same question: split into four 16×16, or keep as one? The recursion continues down to 8×8 or 4×4. The final pattern is a quadtree — each node has either zero or four children — and represents the encoder's content-adaptive partition of the frame.
The benefit is detail-aware coding. Large flat regions (sky, walls, out-of-focus backgrounds) end up as single 64×64 blocks — minimal header overhead, efficient transform, compresses to almost nothing. Detailed regions (eyes, hair, text) end up as many tiny 4×4 or 8×8 blocks — higher overhead per block but accurate detail capture where the eye actually looks. The encoder picks the split level per region using rdo (rate-distortion optimisation) — it tries different splits and keeps the one with the best quality-versus-size trade-off.
For a product team, quadtree partitioning is invisible plumbing that explains a big chunk of HEVC's efficiency win over H.264. H.264 had fixed 16×16 macroblocks; HEVC's quadtree splits to whatever the content demands, and that flexibility alone delivers ~15–20 % of the efficiency gain. AV1 extends quadtree further with non-square partitions (a 64×64 block can split into two 32×64 rectangles instead of four 32×32 squares), and VVC goes further still. Each level of flexibility costs encoder compute but saves bits in the bitstream.
