Filter-based monte carlo gradient path tracing rendering image reconstruction method and device
By obtaining the initial image and auxiliary information map of the Monte Carlo gradient path tracing rendering, and using the gradient and filtering optimization objective function to generate multi-resolution filtering results, the problems of poor image quality and insufficient consistency of the Monte Carlo gradient path rendering method at low sampling numbers are solved, and efficient and stable image reconstruction effects are achieved.
Patent Information
- Application Number
- CN202411503682.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-25
- Publication Date
- 2025-10-10
- Estimated Expiration
- 2044-10-25
AI Technical Summary
Existing Monte Carlo gradient path rendering methods perform poorly at low sampling numbers and cannot guarantee the consistency of calculation results in different scenarios. In particular, the image quality generated by the Poisson equation-based method is unstable at low sampling numbers, and the generalization ability of the neural network method is unstable.
By obtaining the initial rendered image and auxiliary information map obtained by Monte Carlo gradient path tracing rendering, the optimized gradient is solved using the gradient optimization objective function, and multi-resolution levels are generated. Combined with the filtering optimization objective function, filtering calculations are performed layer by layer to generate high-quality and consistent denoised images.
It generates high-quality and consistent denoised images at a lower sampling rate, reduces the impact of noise in gradient estimation, has good generalization performance, and meets the needs of high-efficiency and high-quality image generation.