Face Image Characteristic Point Positioning via Pixel Normalization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current techniques for specifying the positions of characteristic portions in face images, such as Active Appearance Models (AAM), face challenges in achieving accurate positioning due to variations in pixel value distributions, leading to suboptimal accuracy and processing efficiency.
Innovation Solution
An image processing apparatus that includes an initial disposition unit, image transforming unit, normalization unit, determination unit, and update unit, which determines the initial disposition of characteristic points in a target face image by comparing it to a reference face image, performs transformations to align disposition patterns, and updates the disposition based on normalized pixel value indices to improve accuracy and processing speed.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional AAM techniques are used to specify characteristic portions, then the process can be implemented, but the accuracy of position specification is insufficient due to variations in pixel value distributions
Solution Approach 1:
The patent applies parameter changes by normalizing pixel values using statistical parameters (mean and standard deviation) calculated from reference images. This transformation adjusts the pixel value distribution of target images to match the reference distribution, making the comparison invariant to illumination and other environmental variations. The normalization process changes the parameters of pixel value distribution rather than directly comparing raw pixel values, thereby resolving the contradiction between measurement precision and adaptability.
Solution Approach 2:
The patent introduces an intermediary normalization process between image acquisition and characteristic portion detection. This intermediary step transforms the target image pixel values to match the reference image distribution, serving as a mediator that eliminates the adverse effect of distribution variations. The normalization acts as a bridge that allows accurate comparison despite differences in imaging conditions.
2Measurement precision
If multiple transformations and normalizations are performed to improve accuracy, then position specification accuracy improves, but processing time and computational complexity increase
Solution Approach 1:
The patent applies preliminary action by pre-calculating and storing statistical parameters (mean and standard deviation) from reference images before actual processing. This preparation work is done in advance, so that during target image processing, only the normalization transformation is needed rather than full statistical analysis. This reduces the computational burden during time-critical processing while maintaining accuracy.
Solution Approach 2:
The patent transforms the complex problem of accurate position specification under varying conditions into a simpler parameter matching problem. By changing the representation from raw pixel values to normalized parameters, the system achieves high accuracy with fewer computational steps, effectively reducing processing time while maintaining precision.
3Productivity
If characteristic point disposition is updated based on raw pixel value comparison, then the process is simple, but the determination accuracy is low due to pixel value distribution characteristics
Solution Approach 1:
The patent changes the basis of comparison from raw pixel values to normalized parameters. By transforming pixel values to remove distribution variations, the system maintains processing efficiency while dramatically improving determination accuracy. The parameter transformation allows simple comparison operations to yield accurate results.
Solution Approach 2:
The patent replaces the mechanical approach of direct pixel value comparison with a statistical approach using normalized parameters. This substitution transforms a simple but inaccurate mechanical comparison into a more sophisticated statistical comparison that achieves higher accuracy without proportionally increasing complexity.
Data Source
AI summary
There is provided an image processing apparatus that specifies a position of a predetermined characteristic portion of a target face image.