Specularity Classification Using Offset Gaussian Curve Fitting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems for determining specularity of objects from images in non-laboratory settings face inefficiencies due to noisy data and large volumes of images, often resulting in inaccurate or nonsensical results, especially when processing small numbers of images.
Innovation Solution
A computer-implemented method that selects areas of geometry, identifies relevant images, determines intensity and angle values, fits these values to a curve, and classifies specularity using an offset Gaussian curve, allowing for the propagation of shininess between patches and rendering realistic 3D models.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If traditional specularity determination systems process images in non-laboratory settings, then they can handle real-world applications, but the results become noisy and inaccurate due to unknown lighting and surface conditions
Solution Approach 1:
The method segments the problem by separating specularity determination from full BRDF estimation. Instead of attempting to determine all surface properties simultaneously, the invention focuses specifically on extracting specularity parameters (alpha and beta) by analyzing intensity variations with viewing angle, while treating other reflectance components as separable or negligible in the specular direction.
Solution Approach 2:
The invention changes the approach from trying to determine all BRDF parameters to focusing on specific parameters (specularity alpha and beta). By parameterizing the specular component separately and fitting it to intensity data at different viewing angles, the system can determine specularity without needing to solve the entire BRDF equation with all its unknowns.
2Reliability
If systems process large volumes of images to determine specularity, then more data becomes available for analysis, but processing time increases and efficiency decreases
Solution Approach 1:
The method extracts only the necessary information for specularity determination from images - specifically intensity values at different viewing angles - rather than processing and analyzing all image data. By extracting and fitting only the specular component parameters, the system achieves reliable results with fewer images and reduced processing requirements.
3Productivity
If systems attempt to model specularity from limited images (40 or fewer), then processing time is reduced, but the number of unknowns becomes large relative to data points, causing the process to break and produce nonsensical results
Solution Approach 1:
The invention applies partial action by focusing only on determining specularity parameters rather than complete BRDF estimation. This partial approach requires fewer data points and computational resources, enabling reliable results from limited images (40 or fewer) without attempting to solve for all BRDF unknowns simultaneously.
Data Source
Figure 1
Figure 2
Figure 3A~3B
AI summary
Aspects of the disclosure relate generally to determine specularity of an object. As an example, an object or area of geometry 512, 712, 812, 1304 may be selected. A set of images 510, 520, or 620, or 710, 720 or 810, 820 or 1306 that include the area of geometry 512 may be captured. This set of images may be filtered to remove images that do not show the area of geometry well, such as if the area is in a shadow or occluded by another object. A set of intensity values for the area are determined for each image. A set of angle values for each image is determined based on at least a direction of a camera that captured the particular image when the particular image was captured. The set of average intensities and the set of angle values are paired and fit to a curve 636, 840, 1308. The specularity of the area may then be classified based on at least the fit.