Angle-Dependent Anisotropic Filtering for Texture Sampling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing anisotropic filtering techniques result in undesirable distortions and artifacts, such as aliasing, due to the fixed filter size not accounting for the angle-dependent sampling of texture mapped polygons relative to the screen space, leading to inefficient sampling and increased computational workload.
Innovation Solution
Implementing angle-dependent anisotropic filtering, where the filter length is varied based on the angle of the anisotropic major-axis with respect to the screen-space axes, ensuring uniform sampling and minimizing aliasing across all angles by adjusting the effective sampling interval.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a fixed filter size is used in anisotropic filtering, then the filtering process is simple and fast, but it causes undesirable distortions and aliasing artifacts due to angle-dependent sampling
Solution Approach 1:
The filter size is made dynamic by adjusting the major-axis length based on the screen space angle. The filtering process transitions from a static fixed-size filter to a dynamic angle-dependent filter that adapts its dimensions according to the orientation of the texture-mapped polygon relative to the screen space, thereby eliminating aliasing artifacts while maintaining computational efficiency.
Solution Approach 2:
The patent changes the parameter of filter size from a fixed value to a variable that depends on the screen space angle. By modifying the major-axis length according to the angle θ between the texture coordinate system and screen space axes, the filtering process achieves angle-independent sampling that prevents moiré patterns and aliasing while maintaining image quality.
2Reliability
If the filter length is adjusted to account for angle-dependent sampling, then aliasing is reduced, but the computational workload increases
Solution Approach 1:
The filtering process applies different filter dimensions locally based on the specific angle of each texture-mapped polygon. Rather than using a uniformly complex filtering approach for all cases, the patent adjusts the major-axis length only to the extent necessary for each specific angle θ, thereby achieving accurate angle-dependent sampling without unnecessary computational overhead.
Solution Approach 2:
The patent applies partial adjustment to the filter size by modifying only the major-axis length based on the screen space angle, while keeping the minor-axis length constant. This partial action approach achieves the necessary sampling accuracy to prevent aliasing without implementing a fully dynamic two-dimensional filter adjustment, thus maintaining computational efficiency.
3Reliability
If a larger filter size is used to cover all angles, then aliasing is minimized, but unnecessary blur is introduced in axial directions
Solution Approach 1:
The patent introduces asymmetry in the filter dimensions by making the major-axis length angle-dependent while keeping the minor-axis length constant. This asymmetric filtering approach allows the filter to be larger only in the direction necessary to cover the angle-dependent sampling requirements, while maintaining a compact dimension in the perpendicular direction, thereby preventing both aliasing and excessive blur.
Solution Approach 2:
The major-axis length of the filter is made dynamic, varying with the screen space angle θ. When the texture-mapped polygon is aligned with the screen axes (θ = 0° or 90°), the major-axis length reduces to the minimum necessary size, preserving image sharpness. When the polygon is at intermediate angles, the major-axis length increases to prevent aliasing, thus dynamically optimizing the balance between sharpness and aliasing reduction.
Data Source
AI summary
A method includes computing an anisotropic filter with a major-axis and a minor-axis for a pixel to be displayed on screen-space, where the anisotropic filter is to be applied to corresponding MIPs on a texture map. Additionally, the method includes varying the length of the major-axis of the anisotropic filter based on the angle of the major-axis of anisotropy with respect to the screen space. Further, the method includes determining a number of texels from the texture map that are to be sampled in the anisotropic filter based on the length of the modified major-axis. The color of the pixel may be determined based on the texels sampled in the anisotropic filter.