The field of novel view synthesis is rapidly advancing with significant improvements in 3D Gaussian Splatting (3DGS) techniques. Recent developments have focused on optimizing 3DGS for real-time rendering on resource-constrained devices, improving rendering quality, and enhancing the technique's ability to capture complex scenes and effects. Notably, innovations in sorting algorithms, compression methods, and blending techniques have led to substantial gains in performance and fidelity. These advancements have far-reaching implications for applications such as augmented and virtual reality, where high-quality and low-latency rendering are crucial.

Some noteworthy papers in this area include: Neo, which introduces a reuse-and-update sorting algorithm for significant performance gains. TR-Gaussians, which proposes a novel representation for high-fidelity rendering of planar transmission and reflection. Beyond Darkness, which presents a unified framework for low-light novel view synthesis using thermal-supervised 3D Gaussian Splatting. SymGS, which leverages local symmetries for 3DGS compression, achieving substantial compression ratios. Opt3DGS, which enhances 3DGS optimization through adaptive exploration and curvature-aware exploitation. IBGS, which efficiently models fine details and view-dependent effects using image-based Gaussian splatting. Gaussian Blending, which rethinks alpha blending in 3DGS to reduce visual discrepancies. Optimizing 3D Gaussian Splattering for Mobile GPUs, which develops an optimized mapping of 3DGS for mobile GPUs, achieving significant speedups and memory reductions.