2021年12月3日 · We present NeRF-SR, a solution for high-resolution (HR) novel view synthesis with mostly low-resolution (LR) inputs. Our method is built upon Neural Radiance Fields (NeRF) that predicts per-point density and color with a multi-layer perceptron.

Abstract: We present NeRF-SR, a solution for high-resolution (HR) novel view synthesis with mostly low-resolution (LR) inputs. Our method is built upon Neural Radiance Fields (NeRF) that predicts per-point density and color with a multi-layer perceptron.

基于这个缺点本作者提出了Nerf-SR， 能够创造出高质量的SR图像。 任务：采用低分辨率图像训练，产生良好的多视图超分辨率图像。 1. Nyquist view sampling rates 大多数超分方案无法达到这样的采样率。 所以作者提出 超采样策略，以亚像素的方式更好地加强了NeRF中嵌入的多视图一致性。 2.全景图和光场成像系统会在角度分辨率和空间分辨率之间进行权衡，直接将其生成的图像用Nerf进行训练只能在一定范围内提升SR精度。 所以作者提出 patch-wise warp-and-refine …

We present NeRF-SR, a solution for high-resolution (HR) novel view synthesis with mostly low-resolution (LR) inputs. Our method is built upon Neural Radiance Fields (NeRF) \cite {mildenhall2020nerf} that predicts per-point density and color with a multi-layer perceptron.

2022年4月14日 · 本文提出NeRF-SR，一种利用超采样策略和细化网络的解决方案，能在低分辨率输入下生成高质量的高分辨率新视图。 方法基于神经辐射场，通过多视图约束和HR参考图像细节补充，有效提升新视图合成效果。 实验在合成数据集和真实世界场景中展示了卓越性能。 1. Introduction. 5. Experiments. 5.1. Dataset and Metrics. 5.2. Training Details. 5.3. Effectiveness of super-sampling. 5.4. Refifinement network. 6. Limitation. 我们提出了 NeRF-SR，一种高分辨 …

2024年6月30日 · NeRF-SR is the first model to obtain decent multi-view super-resolution results with only low-resolution input images, but the super-sampling method implemented using the original Nerf’s MLP network cannot represent the complex details of the scene well.

2024年6月19日 · nerf-sr是基于神经辐射场（nerf）的一种解决方案，它扩展了nerf的潜力，使其能够处理超出原始观察图像范围的高分辨率问题。通过对每个图像像素发射多条光线，nerf-sr强化了多视图一致性，确保了亚像素级别的信息交换。

2024年2月19日 · 论文提出了一种参考引导的超分神经辐射场 (RefSR-NeRF)，它将 NeRF 扩展到超分辨率和逼真的新颖视图合成。 NeRF 在高分辨率渲染中存在模糊问题，因为其固有的多层感知器难以学习高频细节，并且随着分辨率的增加而导致计算爆炸。

We present NeRF-SR, a solution for high-resolution (HR) novel view synthesis with mostly low-resolution (LR) inputs. Our method is built upon Neural Radiance Fields (NeRF) that predicts per-point density and color with a multi-layer perceptron.

Abstract: We present NeRF-SR, a solution for high-resolution (HR) novel view synthesis with mostly low-resolution (LR) inputs. Our method is built upon Neural Radiance Fields (NeRF) that predicts per-point density and color with a multi-layer perceptron.