The concept originated in the 1970s with adaptive image enhancement systems, where early researchers realized that different parts of an image often required different levels of contrast adjustment or noise suppression. The term is most widely used in Norwegian and some Scandinavian literature, and it has been translated into English as “position-based filtering” or “positional filtering.” In the 1990s, advances in digital image sensors and computing power made real-time position-based filtering practical for applications such as video surveillance, medical imaging and consumer electronics.
Typical algorithms for posisjonsfiltrering include weighted median filters, where the weight of each pixel in a neighborhood varies with distance to the centre, and spatially variant Gaussian blurs, where the kernel size is a function of the pixel’s location. More sophisticated methods use machine‑learning models that output filter parameters conditioned on position, enabling tasks such as edge‑preserving smoothing in photographs and denoising of depth maps.
Applications are numerous. In automotive radar systems, posisjonsfiltrering helps suppress clutter by discarding reflections that arrive from impossible positions on the vehicle. In computer vision, it aids background subtraction by assigning different thresholds to central versus peripheral pixels. In medical imaging, position‑dependent de‑noise filters improve visualisation of structures located in particular anatomical zones.
Despite its utility, posisjonsfiltrering faces challenges. The design of spatially varying parameters can be computationally expensive, especially for high‑resolution data. Moreover, improper parameter choices can introduce artefacts or bias, compromising the integrity of the processed signal. Ongoing research focuses on automatic parameter optimisation and efficient implementation on GPUs, ensuring that position‑based filtering remains a practical tool for modern image‑ and signal‑processing workflows.