Sony Patents a Video Decoding Shortcut That Drops One Neighbor Block
Sony has found a way to speed up video compression hardware by deliberately ignoring one neighboring block of pixels, a small rule change that has real consequences for how quickly chips can process video in a pipeline.
What Sony's motion-vector shortcut actually changes
When your TV, phone, or streaming stick decodes video, it constantly makes educated guesses about how objects are moving from one frame to the next. To make those guesses, the chip looks at blocks of pixels that are nearby in the current frame and asks: where did they come from? The problem is that one particular neighbor, the block sitting to the upper-right of the current one, creates a bottleneck because the chip has to wait for that block to finish before it can move on.
Sony's patent fixes this by simply telling the chip: don't look at that upper-right neighbor at all. Instead, the chip uses only the blocks above and to the left, which are already available earlier in the processing sequence. You get the same prediction, just without the wait.
It sounds like a small bookkeeping change, but in video hardware, pipeline bottlenecks compound fast. Shaving a dependency like this can let the chip process more frames per second or use less power doing the same job.
prohibit, when generating a spatial motion vector predictor for a motion vector predictor used for decoding a motion vector of a current block of an image, use of a motion vector of an above-right block located adjacent to an upper right of the current block; …
Translation: The system bans using motion data from the upper right block to predict the current block's motion.
How the chip skips the upper-right block on purpose
Video compression standards like HEVC and similar codecs rely on a technique called motion vector prediction: instead of storing where every block of pixels moved, the encoder stores only the difference between where it actually moved and where the chip predicted it would move. Smaller differences mean smaller files.
To generate that prediction, the chip looks at spatial neighbor blocks, the blocks immediately surrounding the current one. Traditionally, the standard includes a block called "above-right" (sometimes labeled C0 or B0), the block diagonally adjacent to the upper-right corner of the current block. The catch: that block is processed after the current row in a pipeline, so the chip has to stall and wait for it.
Sony's patent covers circuitry that prohibits the use of that above-right block entirely during two common prediction modes:
- AMVP (Advanced Motion Vector Prediction): where the encoder picks the best predictor from a small candidate list.
- Merge mode: where the encoder copies a neighboring block's full motion information instead of encoding a new vector.
With the above-right block off the table, the circuitry builds its predictor from only four candidates: B1 and B2 (blocks at the top) and A0 and A1 (blocks to the left). All four are available earlier in the pipeline, so the stall disappears and the chip can keep moving without waiting.
… when a spatial prediction motion vector is derived according to AMVP or Merge mode, the use of a motion vector of a PU adjacent to a top right of a PU in question is prohibited.
Translation: During standard prediction modes, utilizing the top right neighbor's motion vector is strictly blocked.
What this means for video chip designers
For most readers, this patent is invisible infrastructure. It lives inside the silicon of a streaming device, a camera, or a TV, and its job is to make sure the chip never sits idle waiting for data it doesn't strictly need. Video chip designers and codec engineers are the people who feel this most directly: removing a pipeline dependency is one of the cleanest ways to raise throughput without adding more transistors or raising clock speed.
The broader context here is a long-running effort by chip and camera makers to squeeze more performance out of fixed hardware budgets, and Sony's codec work is a steady part of the Big Tech patent news that tracks how video compression standards get shaped at the silicon level. Whether this specific rule ends up in a shipping product or influences a future codec profile, it reflects the kind of careful, unglamorous engineering that determines whether your 4K stream stutters or not.
The patent's core rule is simple. When a chip figures out how a video frame moves from one moment to the next, it must ignore one specific surrounding region, the one sitting above and to the right, and rely only on the other surrounding regions instead.
That one deliberate exclusion is narrow enough to hold up in court, but it is also broad enough to cause real trouble. Any company building a video chip that skips that same region for the same practical reason, whatever they call the technique internally, could find itself inside this patent's reach.
Companies designing video chips right now should read this claim closely. The rule about which surrounding region to ignore is the whole point of the protection.
There are more where this came from
We read every patent application Big Tech publishes and send you the ones worth knowing. Plain English, free, every week.
The drawings
29 drawing sheets from US 2026/0246963 A1 · click any drawing to enlarge
Want this weekly breakdown for a company we don't cover? Patentlyze Pro →