Calculates missing pixels during image scaling using cubic functions — sharper than linear, fewer artifacts. Gold standard for high-quality scaling in grading.
When you enlarge a digital image — whether it's upscaling SD material to 2K or zooming into a detail view — the software has to decide what color values the new, added pixels will receive. Bicubic interpolation solves this problem more elegantly than the naive linear method: it considers not only the direct neighboring pixels but works with cubic polynomials over a 4x4 pixel grid. The result is a smoother transition, fewer jaggies, and significantly less aliasing artifacts.
On a practical set, you notice this primarily in grading and post-production. If you want to zoom into a specific area of 4K material, or if you're placing lower-resolution archive footage into a high-resolution project — bicubic is your default choice here in DaVinci Resolve, Nuke, or After Effects. Linear (bilinear) produces visible softness and blockiness; nearest neighbor is reserved for pixel art. Bicubic strikes a balance: sharpness information is preserved without creating artificial edges that can arise with more aggressive algorithms. In editing, you'll especially notice the difference with titles, where edges need to be clean, or with extreme reframing of footage.
Technically, bicubic works with four cubic basis functions per axis — hence the name. Each new pixel value is calculated from sixteen neighbors, weighted by their distance. This is more computationally intensive than bilinear, but no longer a real problem on modern hardware. Some systems offer variants: Catmull-Rom splines, Mitchell filters, or even higher-order approaches for extreme enlargements. However, for everyday VFX work, bicubic is perfectly sufficient and considered the gold standard because the computation time is minimal and the visual result is reliable.
A practical tip: with multiple scaling operations — first during proxy creation, then again in final rendering — you should keep bicubic consistent. Mixing different interpolation methods on layers or during resampling leads to subtle quality losses. And if your grading system offers a choice: bicubic-sharp can sometimes look better than standard bicubic, but it also pulls out artifacts more sharply. Previewing is essential here.