Face Classification Using Regional Segmentation for Expression Robustness

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing face classification apparatuses struggle to accurately detect faces in images with varying facial expressions and orientations, as they are influenced by changes in facial angles and directions, leading to lower detection accuracy in general images compared to controlled environments like photographs or authentication images.

Innovation Solution

A face classification method using a machine-learning approach that learns characteristic features from a diverse set of facial images with specific directions and angles, focusing on predetermined facial regions such as entire facial contours, eyes, noses, and upper lips, to improve detection accuracy and robustness across different orientations and expressions.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If a face classification apparatus learns from facial images with varying directions and angles, then the adaptability to different orientations improves, but the detection accuracy deteriorates due to the complexity introduced by diverse facial expressions and angles

Engineering Contradiction:

Improveadaptability to different orientationsVSAvoiddetection accuracy

Core Design Contradiction:

Adaptability or versatilityVSMeasurement precision

Solution Approach 1:

The patent segments facial images into multiple predetermined regions (eyes, nose, mouth, cheeks, forehead) and trains separate classifiers for each region. This segmentation allows the system to handle varying directions and angles more effectively by focusing on specific facial features that remain relatively stable across different orientations, thereby maintaining detection accuracy while improving adaptability.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent applies local quality by training different classifiers with different characteristics for different facial regions. Each classifier is optimized for its specific region, allowing the system to adapt to local variations in facial expressions and orientations while maintaining overall detection accuracy. This regional specialization enables the apparatus to handle diverse orientations without sacrificing precision.

Inventive Principle:

Principle #3Local quality

2Measurement precision

If multiple predetermined regions are extracted and classified separately, then the detection accuracy improves, but the device complexity increases due to multiple classifiers and processing steps

Engineering Contradiction:

Improvedetection accuracyVSAvoidcomplexity of classification system

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The patent divides the face classification task into multiple segments, extracting and classifying different predetermined regions (eyes, nose, mouth, cheeks, forehead) separately. This segmentation improves detection accuracy by focusing on specific facial features, and the modular structure of separate classifiers actually simplifies the overall system design compared to a single complex classifier.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent combines multiple region-based classification results to form an overall face classification decision. By merging the outputs of individual region classifiers through a combination logic, the system achieves high detection accuracy while maintaining a manageable level of complexity through structured integration rather than monolithic processing.

Inventive Principle:

Principle #5Merging (Combining)

3Measurement precision

If facial images with the same direction and angle are used for learning, then the detection accuracy for controlled images improves, but the adaptability to general images with varying orientations deteriorates

Engineering Contradiction:

Improvedetection accuracy for controlled imagesVSAvoidadaptability to general images

Core Design Contradiction:

Measurement precisionVSAdaptability or versatility

Solution Approach 1:

Inventive Principle:

Principle #3Local quality

Solution Approach 2:

The patent creates a universal face classification system that can handle multiple types of images (controlled photographs and general snapshots) by using a multi-functional classifier structure. The same classification framework processes both controlled images with consistent orientations and general images with varying orientations, achieving both high accuracy for controlled images and good adaptability for general images.

Inventive Principle:

Principle #6Universality (Multi-functionality)

Data Source

PatentUS7835549B2Learning method of face classification apparatus, face classification method, apparatus and program

Publication Date: 2010.11.16 FUJIFILM CORP

AI summary

A plurality of different facial images is used to cause a face classification apparatus to learn a characteristic feature of a face by using a machine-learning method. Each of the facial images includes a face which has the same direction and the same angle of inclination as those of a face included in each of the other facial images and each of the facial images is limited to an image of a specific facial region. For example, the facial region is a predetermined region including only a specific facial part other than a region below an upper lip to avoid an influence of a change in facial expressions. Alternatively, if the apparatus is used to detect a frontal face and to perform refined detection processing on the extracted face candidate, a region including only an eye or eyes, a nose and an upper lip is used as the facial region.