ꟻLIP‐Based Perceptual Optimization for Subpixel Rendering Filters

ABSTRACT We integrate ꟻLIP into an MMSE‐based subpixel‐rendering framework to improve display visual performance. Using mean ꟻLIP errors as loss in a Nelder–Mead optimizer, we refine the kernel to reduce perceptual distortion. The proposed method begins with an MMSE‐derived kernel based on a virtual‐image reconstruction model, which provides a physically meaningful baseline filter for subpixel rendering. The perceptual optimization stage then iteratively adjusts the kernel coefficients to minimize ꟻLIP errors rather than purely numerical RGB differences, enabling improved chromatic accuracy and edge fidelity. Verified with ISO 12233, DIV2K, TID2013, and ImageNet, the method achieves a 27.67% reduction in mean ꟻLIP errors over the diagonal direct subpixel‐based down‐sampling approach, based on a mini‐LED panel with GRBG subpixel layout. These results demonstrate that perceptually guided subpixel rendering provides an effective strategy for improving perceptual image quality in modern display systems.