High‐Fidelity Ultra‐High‐Definition Chroma Reconstruction in the Scaled YCbCr Colour Space via LUL ₀ Factorisation and Generalised Lifting

Chroma subsampling reduces bandwidth and storage requirements in digital image processing by encoding chroma components at a lower spatial resolution than luminance. When implemented in the conventional YCbCr colour space, however, this process may perturb the geometry of the underlying colour transform, which can degrade chroma reconstruction fidelity after interpolation. This paper proposes a scaled YCbCr colour space to improve ultra‐high‐definition chroma reconstruction accuracy in high‐resolution imaging scenarios. The proposed framework comprises three key components: (i) constructing a diagonally scaled YCbCr transform with unit determinant to preserve colour‐space volume; (ii) analytically factorising the scaled transform into three triangular matrices, referred to as the LUL 0 factorisation; and (iii) realising integer‐to‐integer forward and inverse transforms via a generalised lifting scheme derived from this factorisation. In our experimental setting, the proposed scaled YCbCr colour space demonstrates improved preservation of chromatic relationships relative to the conventional YCbCr colour space, thereby reducing interpolation‐induced distortions. On the DIV2K validation images with 4:2:0 chroma subsampling, the proposed method achieves the highest PSNR among the evaluated chroma prediction methods published between 2020 and 2024, including both learning‐based and non‐learning‐based approaches. Comprehensive evaluation using multiple metrics (colour peak signal‐to‐noise ratio, CIEDE2000, structural similarity index, feature similarity index, and mean absolute error) and additional benchmark datasets (DIV2K, Kodak, BSD, and the screen‐content set SCID) further confirms the robustness of the proposed approach.