图像渲染模型训练、图像渲染方法及装置
By using a combination of explicit density distribution matrix and voxel sampling in NeRF image rendering, the problem of high computational cost and low efficiency is solved, resulting in faster rendering speed and higher efficiency.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- SHANGHAI BIREN TECH CO LTD
- Filing Date
- 2022-01-17
- Publication Date
- 2026-07-17
AI Technical Summary
Existing NeRF image rendering methods are computationally intensive and inefficient, especially during network training, which results in slow rendering speeds.
By determining the target scene map from multiple angles, an explicit density distribution matrix is obtained through projection reconstruction. The initial neural radiation field is trained by combining voxel sampling and density difference, reducing the number of voxel samplings. The density information is directly obtained by using the explicit density distribution matrix, reducing forward computation.
Without increasing the amount of additional computation, the training speed of the image rendering model and the convergence of the loss function are accelerated, thus improving rendering efficiency.
Smart Images
Figure CN114493995B_ABST