bikubische

Bikubische interpolation, commonly called bicubic interpolation in English, is a resampling method used to resize digital images. It estimates new pixel values by fitting a cubic polynomial to the 16 pixels surrounding the target location, resulting in smoother images than bilinear or nearest-neighbor methods. The technique is widely used in image processing, video scaling, and texture filtering in computer graphics, where intermediate image data is required when changing resolution or aspect ratio.

How it works: To determine a new pixel at position (x, y) in the destination image, bicubic interpolation considers the four source pixels along each axis (a 4x4 window) and applies a cubic convolution kernel to interpolate first along one axis and then along the other. The kernel is piecewise cubic and is defined by a parameter a that controls the shape of the curve. Common implementations use Keys' cubic convolution with a around -0.5 to -0.75, producing different smoothing versus sharpness tradeoffs.

Advantages and limitations: Bicubic interpolation yields smoother results with fewer artifacts than bilinear interpolation and preserves more detail in gently varying regions. It can, however, produce slight overshoot or ringing near sharp edges and high-contrast transitions. It is more computationally intensive than bilinear, and may blur fine details in some images. It is not ideal for binary images, such as text, where crisp edges are preferred.

History and usage: Developed as part of the family of polynomial-based resampling methods in the late 20th century, bicubic interpolation became a standard option in image editors, video codecs, and graphics libraries. It remains a default choice in many software packages and hardware implementations due to its balance of quality and performance.