upscaling

Upscaling is a process in digital imaging and video processing that increases the apparent resolution of an image or video frame. It operates by generating new pixel values to fill a larger grid, allowing low-resolution media to be viewed at higher display resolutions or stored at higher formats. Upscaling covers both spatial resolution increases and, in some contexts, frame rate enhancement, though the term is most often used for resolution.

A range of techniques is used. Simple interpolation methods such as nearest-neighbor, bilinear, and bicubic estimate new pixels from surrounding ones. More advanced approaches include Lanczos resampling and edge-directed interpolation to preserve sharpness. Modern systems increasingly use machine learning-based upscaling that trains models to reconstruct plausible high-frequency details from low-resolution input. These methods can deliver impressive results but require more computation and may introduce artifacts if the input is noisy or heavily compressed.

Limitations and evaluation: Upscaling cannot recover true original detail; the quality depends on the method and content. Artifacts such as halos, ringing, or oversharpening can occur. Quality is assessed with objective metrics like PSNR or SSIM, along with subjective visual evaluation. In practice, hardware accelerates real-time upscaling on televisions, GPUs, and video devices.

Applications: Upscaling is widely used to display older or lower-resolution media on high-resolution screens, to fit content to modern formats in streaming and broadcasting, in video production, surveillance, and gaming. It sits alongside image enhancement and the broader field of super-resolution.