Samsung Patents Technology That Keeps Moving Images Sharp Across Every Color Layer

By Patentlyze Team · Updated Jun 26, 2026

Every time a video plays on your phone, a decoder is racing to reconstruct millions of color blocks per second. Samsung's latest patent targets a specific inefficiency in how color motion is calculated, borrowing data from the brightness channel to do it faster.

How Samsung's color-channel borrowing trick works in video

When video is compressed and then decoded, the decoder has to figure out where every part of the image moved between frames. Modern codecs split each frame into tiny blocks and track their motion separately. But color information (called chroma) and brightness information (called luma) are stored differently, and that creates extra work.

Samsung's patent describes a method where, instead of computing motion separately for each color block, the decoder averages the motion vectors already calculated for two neighboring brightness blocks and uses that average to guide the corresponding color block. It's a bit like navigating by combining two friends' directions rather than drawing a new map from scratch.

This matters most when the decoder is using a mode called affine prediction, which handles objects that rotate or zoom rather than just slide. The approach could reduce the computation needed to decode color in those situations, which could translate to smoother playback or lower processing power on devices.

How luma motion vectors drive chroma sub-block prediction

The patent targets affine inter prediction, a technique used in modern video codecs (like VVC/H.266) where motion is described not as a single direction but as a transformation that can handle rotation, zooming, and shearing. To do affine prediction, a decoder needs multiple control point motion vectors (reference points that define how a block is warping, not just translating).

In standard formats, the image is split into separate planes: a high-resolution luma (brightness) plane and lower-resolution chroma (color) planes. Because chroma blocks are physically smaller in the encoded bitstream (often half the width or height of luma blocks), there is no clean one-to-one mapping between a chroma sub-block and a single luma sub-block.

The patent's specific fix: when processing an N×N chroma sub-block, the decoder takes the motion vectors from the first N×N luma sub-block to its left and the adjacent N×N luma sub-block to its right, averages them, and uses that averaged vector as the chroma motion vector. The chroma sub-block is then treated as two halves (N/2×N each) that together span the same spatial area as both luma sub-blocks.

• Obtain control point motion vectors for the coding unit in affine mode
• Derive per-luma-sub-block motion vectors from those control points
• Average two horizontally adjacent luma vectors to get one chroma vector
• Use that chroma vector against a reference picture to generate the predicted color samples

What this means for video compression efficiency

Affine prediction is computationally expensive, and the color channels add to that cost. By deriving chroma motion from already-computed luma vectors rather than running a separate calculation, this approach reduces redundant work inside the decoder. For devices like phones and TVs that decode 4K or 8K video in real time, that kind of efficiency has a direct effect on battery life and heat.

This is incremental, standards-level codec work rather than a product feature anyone will see directly. Samsung is a major contributor to video coding standards bodies, so patents like this often feed into future codec specifications used industry-wide, not just in Samsung products.

Editorial commentary on a publicly published patent application. Not legal advice.