Polarization Image Processing for Surface Shape Identification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing technologies face challenges in accurately determining the surface shape of objects from polarization images due to the periodicity of 180 degrees in polarization direction and luminance, leading to uncertainty in azimuth angle measurement and inability to identify surface shape from two polarization directions.
Innovation Solution
An image processing device and method that acquires polarization images from multiple directions, computes image feature quantities based on normalized luminance, and incorporates non-polarization image features, using gradient strength and direction, to overcome the limitations of single polarization direction processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If polarization images of two polarization directions are used, then surface material identification is enabled, but surface shape identification becomes theoretically impossible
Solution Approach 1:
The patent transitions from using only polarization direction information to incorporating spatial gradient information (another dimension) by computing gradient strength and gradient direction from polarization images. This allows the system to overcome the limitation of two polarization directions by adding spatial derivative information that provides additional constraints for surface shape recovery.
Solution Approach 2:
The patent changes the parameters used for surface characterization by computing not only intensity and polarization angle but also gradient strength and gradient direction. This parameter expansion transforms the insufficient two-directional polarization data into sufficient information for both surface material and surface shape identification.
2Reliability
If normalization is applied to luminance in feature quantity computation, then robustness to illumination differences is improved, but computation complexity increases
Solution Approach 1:
The patent performs normalization of luminance values as a preliminary step before computing gradient strength and gradient direction. By pre-normalizing the luminance data, the system ensures that subsequent gradient computations are based on illumination-invariant features, improving robustness while organizing the computation flow to manage complexity systematically.
3Measurement precision
If gradient strength and gradient direction are computed from non-polarization images, then feature quantity accuracy is improved, but processing time increases
Solution Approach 1:
The patent merges the computation of gradient strength and gradient direction directly into the polarization image processing pipeline. Instead of computing gradients from non-polarization images as a separate step, the system computes these gradient features directly from the polarization image data, combining multiple operations into a unified process that reduces processing time while maintaining accuracy.
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 precise identification of object surface shapes by computing image feature quantities that are robust to illumination differences and capable of distinguishing surface materials, effectively addressing the uncertainty of 180 degrees and enhancing object recognition and feature point detection.
Implementation Method 1
an intensity of light of each polarized light component is measured by separating reflected light from the reference surface into s-polarized light and p-polarized light
Implementation Method 2
separating reflected light from the reference surface into s-polarized light and p-polarized light
Data Source
AI summary
A polarization image acquisition unit (11) acquires polarization images of three or more polarization directions. A feature quantity computation unit (15) computes image feature quantities on the basis of the acquired polarization images. For example, the luminance of each polarization image is normalized for each pixel, and the normalized luminance of the polarization image is used as the image feature quantity. The luminance of the polarization image changes according to the surface shape of an object. Thus, the image feature quantities computed on the basis of the polarization images are feature quantities corresponding to the surface shape of the object. Image processing, for example, image recognition, feature point detection, feature point matching, or the like, can be performed on the basis of the surface shape of the object using such image feature quantities.