Introducing a Radial Transform and its Application in Video Compression

Abstract

In this paper, a novel transformation inspired by the patterns in field dynamics is proposed. This transformation, coined as `Radial Transform’ has far-reaching effects and can capture a whole slew of transformations. The core idea behind the spherical construct of the Radial Transform is to exploit the rotational invariance present in signals in the transformed domain, unlike the conventional Discrete Cosine Transform (DCT), which exploits the translational invariance of sinusoidal waves. The Radial Transform is employed within a video codec as a replacement for the DCT. The authors’ results demonstrate that the Radial Transform achieves an increase in Peak Signal-to-Noise Ratio (PSNR) performance of 1.5 dB compared to DCT while maintaining the entropy in the regime around 0.7 and a minute difference in Multi-Scale Structural Similarity Index (MSSSIM). In texture analysis, superior compression results for texture-rich images and competitive performance in intricate patterns are observed. Incorporating the Radial Transform into the video codec presents a compelling proposition with advantages that significantly outweigh any associated disadvantages. This endeavor, initially inspired by field dynamics, has evolved into a novel and promising innovation.

Paper