用NeRF隐式建模解决遥感图像拼接色差问题
A Nerf-Based Color Consistency Method for Remote Sensing Images
- 用NeRF隐式融合多视角图像特征，统一光照表达
- 在超视界-1和无人机数据上实现无缝色彩过渡
- 适合遥感影像拼接与跨时相图像融合场景
由于季节、光照和大气条件差异，获取的影像光度变化显著，导致拼接图像边缘出现明显接缝。传统方法分为绝对辐射校正和相对辐射归一化两类。本文提出一种基于NeRF的多视角图像颜色一致性校正方法，通过隐式表达融合图像特征，并重光照特征空间，生成具有新视角的融合图像。实验选用具有大范围和时间差异的Superview-1卫星图像与无人机图像进行验证。结果表明，该方法生成的合成图像视觉效果优异，边缘色彩过渡平滑。
原文摘要 · Abstract (English)
Due to different seasons, illumination, and atmospheric conditions, the photometric of the acquired image varies greatly, which leads to obvious stitching seams at the edges of the mosaic image. Traditional methods can be divided into two categories, one is absolute radiation correction and the other is relative radiation normalization. We propose a NeRF-based method of color consistency correction for multi-view images, which weaves image features together using implicit expressions, and then re-illuminates feature space to generate a fusion image with a new perspective. We chose Superview-1 satellite images and UAV images with large range and time difference for the experiment. Experimental results show that the synthesize image generated by our method has excellent visual effect and smooth color transition at the edges.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。