XScale-NVS: Cross-Scale Novel View Synthesis with Hash Featurized Manifold

arXiv:2403.19517v1 Announce Type: new
Abstract: We propose XScale-NVS for high-fidelity cross-scale novel view synthesis of real-world large-scale scenes. Existing representations based on explicit surface suffer from discretization resolution or UV distortion, while implicit volumetric representations lack scalability for large scenes due to the dispersed weight distribution and surface ambiguity. In light of the above challenges, we introduce hash featurized manifold, a novel hash-based featurization coupled with a deferred neural rendering framework. This approach fully unlocks the expressivity of the representation …

abstract arxiv cs.cv distribution fidelity hash light manifold novel resolution scalability scale surface synthesis type view world