Guided Filter Denoising for Path-Traced 3D Images
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Path-tracing in computer graphics is computationally expensive due to the noise inherent in the method, making it impractical to achieve high-quality images with low noise without increasing the number of samples.
Innovation Solution
A method is provided that uses guided filtering to denoise path-traced images by rendering noisy images at a first resolution, obtaining guide channels at the same or higher resolution, and calculating model parameters to approximate the noisy image as a function of the guide channels, producing a denoised image at the higher resolution.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the number of samples per pixel is increased to reduce noise in path-traced images, then the image quality improves, but the computational cost increases significantly
Solution Approach 1:
The patent introduces guide channels as intermediary data structures that capture scene structure information (depth, surface normals, ambient occlusion) separately from the noisy rendered image. These guide channels serve as mediators that enable the guided filter to denoise the image by referencing the structure information, avoiding the need to increase sample count while maintaining image quality.
Solution Approach 2:
The patent segments the rendering process into separate components: rendering the noisy image at reduced sample count, rendering guide channels with structure information, and applying guided filtering. This segmentation allows each component to be optimized independently, reducing overall computational cost while maintaining quality.
2Measurement precision
If the number of samples per pixel is increased to reduce noise, then the noise level decreases, but the rendering time increases
Solution Approach 1:
The patent performs preliminary rendering of guide channels that contain structure information before the final image rendering. By preparing these guide channels in advance with appropriate structure data, the guided filter can efficiently denoise the final image without requiring excessive samples, thereby reducing total rendering time.
3Measurement precision
If guided filtering is applied to denoise the noisy image, then the noise is reduced without significant blurring, but additional computational resources are required
Solution Approach 1:
The guided filter operates locally by computing affine transformations for each pixel based on its neighborhood, using guide channels that represent local scene structure. This local approach allows the filter to adapt to local variations in scene geometry and lighting, achieving high denoising quality without excessive global computation.
Data Source
AI summary
An image of a 3-D scene is rendered by first rendering a noisy image at a first resolution. One or more guide channels at the first resolution and one or more corresponding guide channels at a second resolution are obtained. When the two resolutions are the same, the guide channels at the first resolution and the corresponding guide channels at the second resolution may be provided by a single set of guide channels. For each of a plurality of local neighborhoods, the parameters of a model that approximates the noisy image as a function of the one or more guide channels (at the first resolution) are calculated, and the calculated parameters are applied to the one or more guide channels (at the second resolution), to produce a denoised image at the second resolution. The one or more guide channels include at least one guide channel characterizing a spatial dependency of incident light on global lighting over the surface of one or more 3-D models in the scene.