Photometric Stereo Normal Estimation With Smooth Reliability Weighting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing techniques for deriving normal information from objects suffer from inaccuracies due to shadows or specularly reflected light, leading to incorrect relationships between light angles and reflection, and artifacts in output normal information.
Innovation Solution
An image processing apparatus that sets low reliability areas in captured images to identify regions affected by shadows or specular reflections, applies weights that smoothly transition between reliable and unreliable areas, and uses photometric stereo methods to derive accurate normal information.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If image processing is performed using captured images with shadows or specular reflections, then normal information can be derived, but the accuracy of normal derivation deteriorates due to incorrect light reflection relationships
Solution Approach 1:
The captured image is segmented into multiple pixel groups based on luminance values. By dividing the image into regions with different reliability levels (high reliability pixel groups and low reliability pixel groups), the system can process reliable regions with standard algorithms while handling unreliable regions separately, thus improving overall normal derivation accuracy despite the presence of shadows or specular reflections
Solution Approach 2:
Different processing strategies are applied to different regions of the image based on their local quality characteristics. High reliability pixel groups undergo standard normal derivation processing, while low reliability pixel groups are identified and excluded or specially handled. This local differentiation ensures that accurate regions contribute maximally to the result while unreliable regions (shadows, specular highlights) do not degrade the overall accuracy
2Measurement precision
If image processing is performed using captured images with shadows or specular reflections, then normal information can be derived, but artifacts occur in the output normal information
Solution Approach 1:
Low reliability pixel groups that cause artifacts (such as those in shadow regions or specular reflection areas) are extracted and identified separately from the main processing flow. By isolating these problematic regions through luminance-based classification, their harmful influence on the overall normal map is eliminated, preventing artifact generation in the output
Solution Approach 2:
The presence of shadows and specular reflections, which normally create artifacts, is converted into a benefit by using luminance value analysis to identify and separate these regions. The harmful effect is transformed into a useful classification mechanism that allows the system to automatically detect and handle problematic areas, improving the overall quality of the normal information output
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
Enables high-accuracy derivation of normal information even in the presence of shadows or specular reflections by setting low reliability areas and applying spatially smooth weights, preventing artifacts and improving accuracy.
Implementation Method 1
capturing an image of an object to which light is emitted from a plurality of lighting units
Data Source
AI summary
An image processing apparatus includes an acquisition unit configured to acquire image data obtained by capturing an image of an object to which light is emitted from a plurality of lighting units, a setting unit configured to set a first area with a low reliability for deriving normal information regarding the object in an image represented by the acquired image data, a first derivation unit configured to derive a weight for deriving the normal information regarding the object so that the weight continuously changes between the first area and a second area different from the first area, and a second derivation unit configured to derive the normal information regarding the object based on the derived weight.