Spectral Alpha Matting for Garment and Mannequin Separation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing techniques fail to accurately generate alpha mattes for foreground objects, such as garments on a mannequin, due to the lack of physical separation between the mannequin and the garment, leading to issues like color spill and background shadow, and are not suitable for semi-transparent garments or accurate body model generation.
Innovation Solution
A method and system that measure the attenuation of light between points on the mannequin or backdrop and the camera, allowing for the generation of alpha mattes with a background that includes foreground objects, using different spectral power distributions for image acquisition to distinguish between the mannequin and garments, and a process to generate accurate body models from measurements or images.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If conventional alpha matte generation techniques are used on garments on a mannequin, then the process can be automated, but the accuracy of separation between garment and mannequin deteriorates due to lack of physical separation and color spill
Solution Approach 1:
The patent segments the imaging process into multiple spectral acquisitions, capturing images at different wavelengths (e.g., visible and infrared). This allows separate characterization of garment and mannequin spectral signatures, enabling accurate alpha matte computation even without physical separation between foreground and background objects.
Solution Approach 2:
The patent adds a spectral dimension to the traditional spatial imaging approach. By acquiring images across multiple spectral bands and analyzing spectral power distributions, the system distinguishes between garment and mannequin based on their different spectral characteristics rather than relying on spatial separation.
2Measurement precision
If multiple spectral acquisitions are performed to distinguish foreground objects, then separation accuracy improves, but the time required for image processing increases
Solution Approach 1:
The patent performs preliminary spectral characterization by acquiring reference images of the mannequin without garments at multiple wavelengths. These reference spectral signatures are stored and reused for subsequent garment imaging, eliminating the need to repeatedly acquire mannequin references and reducing total processing time.
Solution Approach 2:
The patent replaces time-consuming manual or iterative separation processes with automated spectral analysis algorithms. The system computes alpha mattes by comparing spectral power distributions using programmed algorithms, significantly reducing processing time compared to manual adjustment methods.
3Ease of manufacture
If the mannequin is treated as background, then simple alpha matting can be applied, but color spill from the mannequin contaminates the garment alpha matte
Solution Approach 1:
The patent changes the parameter used for separation from spatial position to spectral characteristics. By analyzing differences in spectral power distribution between garment and mannequin across multiple wavelengths, the system identifies and excludes color spill contamination, treating the mannequin as a distinct spectral signature rather than simple background.
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 automatic computation of alpha mattes with reduced user intervention, accurate separation of foreground and background, and improved realism in image compositing, especially for semi-transparent garments, while allowing for precise body model generation and alignment.
Implementation Method 1
emitting or retroreflecting electromagnetic radiation having a first spectral power distribution from a surface of a first foreground object
Implementation Method 2
emitting or retroreflecting electromagnetic radiation having a first spectral power distribution from a surface of a first foreground object
Implementation Method 3
measuring the attenuation of the garment fabric of light between points on the surface of the mannequin or backdrop and the camera
Data Source
AI summary
Images of foreground objects in a scene are generated by causing electromagnetic radiation to be emitted having a first spectral power distribution from a surface of a first foreground object, which is adjacent or at least partially obscured by a second foreground object. A first image of both of the first and second foreground objects is acquired while the first foreground object emits electromagnetic radiation with the first spectral power distribution. A second image of the first and second foreground objects is acquired while the first foreground object is not emitting electromagnetic radiation or is emitting electromagnetic radiation with a second spectral power distribution which is different to the first spectral power distribution. An alpha matte of the first and second foreground objects is generated based on a comparison of the first image and second image.