A new preprint on arXiv presents a unified theory of image scaling, aiming to connect how images are represented at different resolutions. The authors derive a system of operators, denoted K_s, that they say reproduces optical zoom—specifically, exact zoom-out and canonical zoom-in. The work is framed around a Huygens-like consistency principle, borrowing the idea from wave optics that each point on a wavefront acts as a source of secondary waves. Here, each pixel is treated analogously, providing a physical consistency condition for how scaling operators should behave.
The theory ties together three concepts that are usually treated separately: point spread functions, discrete convolution kernels, and scale-space representations. By aligning these under one derivation, the authors aim to give a principled basis for choosing scaling methods in image processing. The abstract does not include experimental comparisons or implementation details, so the practical advantages over existing methods are not yet demonstrated.
As this is a single preprint, there is no independent source to corroborate or contrast the claims. The contribution is primarily theoretical, offering a new way to think about image scaling rather than a specific algorithm or benchmark result. Readers should treat the findings as preliminary until peer review or follow-up work appears.