Two-Direction Recursive Pixel Filtering for Real-Time Path Tracing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing path and ray tracing techniques in graphical processing systems require a large number of calculations to reduce noise in rendered images, which increases rendering time, especially on mobile platforms with limited processing resources.
Innovation Solution
A pixel filter that applies recursive filtering operations in two directions, combining filtered pixel values to smooth noise while maintaining image quality, using measures of similarity and confidence in pixel values to enhance the filtering process.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If path or ray tracing techniques are used to render images, then image realism and quality are improved, but the number of calculations required increases significantly
Solution Approach 1:
The patent segments the image rendering process into distinct phases: initial path/ray tracing to generate base pixel values, followed by separate denoising operations. This segmentation allows the computationally intensive tracing to be performed once, with subsequent lighter-weight filtering steps to achieve final quality, thereby reducing total calculation requirements while maintaining image realism.
Solution Approach 2:
The patent applies preliminary denoising filters to pixel values before final image composition. By performing denoising operations on intermediate pixel data rather than on the final composite image, the system reduces calculation complexity early in the pipeline, preventing noise propagation through subsequent rendering stages and reducing the overall computational burden.
2Manufacturing precision
If the number of samples per pixel is increased to reduce noise, then image quality is improved, but rendering time increases
Solution Approach 1:
The patent applies denoising filters that operate on pixel values with fewer samples rather than requiring excessive sampling. The filtering process compensates for limited samples by borrowing information from neighboring pixels and temporal data, achieving acceptable image quality with partial sampling (fewer rays per pixel) rather than requiring exhaustive sampling, thereby reducing rendering time while maintaining sufficient quality.
Solution Approach 2:
The patent implements temporal denoising that continuously accumulates and processes pixel values across multiple frames. By maintaining a continuous stream of pixel data and applying recursive filtering that incorporates historical information, the system achieves smooth image quality improvement over time without requiring all samples to be computed simultaneously, effectively parallelizing the quality improvement process across time to reduce perceived rendering time.
3Manufacturing precision
If denoising filters are applied to reduce noise in rendered images, then image quality is improved, but computational overhead increases
Solution Approach 1:
The patent segments the denoising process into multiple passes with different filter strengths and scopes. Instead of applying a single heavy denoising operation, the system uses preliminary lightweight filters followed by more intensive filtering only where needed, reducing overall computational overhead while achieving the same image quality improvement through distributed, modular processing.
Solution Approach 2:
The patent applies denoising filters with spatially varying parameters, adjusting filter strength and radius based on local image characteristics such as edges, textures, and noise levels. By concentrating computational resources only in regions where denoising is most beneficial rather than uniformly processing the entire image, the system reduces total computational overhead while maintaining image quality in critical areas.
Data Source
AI summary
A pixel filter has a filter module that performs a first recursive filter operation in a first direction through a sequence of pixels to form a first filtered pixel value for each pixel, and performs a second recursive filter operation in a second direction through the sequence of pixels to form a second filtered pixel value for each pixel, the first and second recursive filter operations forming a respective filtered pixel value for a given pixel in dependence on the pixel value at that pixel and the filtered pixel value preceding that pixel in their respective direction of operation. The filtered pixel value of the preceding pixel is scaled by a measure of similarity between data associated with that pixel and its preceding pixel. Filter logic combines the first and second filtered pixel values formed by the first and second recursive filter operations to generate a filter output for the pixel, for each pixel of the sequence.