Technology
Converting lower-resolution video to 4K relies on interpolation and AI predictions
Expanding video to 4K resolution requires generating millions of new pixels, leaving software to estimate missing detail without restoring true lost data.
The short version
- Upscaling 1080p video to 4K requires creating three additional pixels for every original pixel to fill a 3,840 x 2,160 frame.
- Traditional tools rely on mathematical interpolation to fill gaps, while newer AI software predicts sharper features based on trained image datasets.
- Neither method can recover unrecorded detail or fix low-bitrate compression artifacts, making original source quality the primary limit on final visual results.
Key facts
- Converting a 1080p video frame to a standard 4K frame requires generating three new pixels for every original pixel, quadrupling the frame's total pixel count.[Engadget]
- Traditional upscaling tools calculate new pixels through mathematical interpolation, which often produces softer image details.[Engadget]
- AI-based upscaling software uses models trained on image pairs to predict sharper details, though it can introduce unique visual artifacts.[Engadget]
- CBS spent over $12 million across three years using 25,000 reels of original 35mm film negatives to remaster Star Trek: The Next Generation into high definition.[Engadget]
- Upscaling expands the resolution grid but cannot resolve underlying compression issues such as low bitrates, blockiness, or color banding.[Engadget]
What remains uncertain
- The extent to which future AI model developments will eliminate visual artifacts and improve output consistency remains uncertain.[Engadget]