Sequential Face Detection Using Gradient Vectors

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing methods for detecting target objects in digital images, such as faces with specific characteristics, are inefficient due to the need for extensive calculations and are not robust against variations in size, orientation, and deformation, leading to inaccurate detection and increased operator burden in automatic ID photograph creation systems.

Innovation Solution

A target object detecting method and apparatus that performs a standard target object detection process followed by specific characteristic detection processes sequentially, using gradient vectors and quaternarized/ternarized characteristic amounts to identify faces with standard and specific characteristics without requiring excessive calculations, allowing for accurate detection of both standard and uniquely featured faces.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Ease of manufacture

If template matching method is used for target object detection, then the detection process is simple, but it cannot tolerate variations in size, direction, and deformation of target objects

Engineering Contradiction:

Improvedetection process simplicityVSAvoidtolerance to object variations

Core Design Contradiction:

Ease of manufactureVSAdaptability or versatility

Solution Approach 1:

The detection process is divided into multiple stages: initial template matching followed by refined detection of remaining objects. This segmentation allows the system to handle variations by processing objects in stages, with each stage addressing specific types of variations.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The system dynamically adjusts detection parameters and methods based on the characteristics of detected objects. By switching between different detection approaches (template matching, characteristic-based detection) and adjusting parameters like threshold values and search regions, the system adapts to handle various object variations effectively.

Inventive Principle:

Principle #15Dynamics

2Difficulty of detecting and measuring

If KL expansion or Hough's conversion is used to project digital image, then target object characteristics are handled more easily, but the methods cannot fully tolerate variations of target object

Engineering Contradiction:

Improvecharacteristic handling easeVSAvoidtolerance to object variations

Core Design Contradiction:

Difficulty of detecting and measuringVSAdaptability or versatility

Solution Approach 1:

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

Inventive Principle:

Principle #15Dynamics

3Reliability

If neural network method is used for robust detection, then detection robustness is improved, but calculation complexity and processing time increase significantly

Engineering Contradiction:

Improvedetection robustnessVSAvoidprocessing time

Core Design Contradiction:

ReliabilityVSLoss of time

Solution Approach 1:

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The system applies partial neural network processing only to regions or objects that require it, rather than processing the entire image through the full neural network. This selective application reduces overall processing time while maintaining robustness where needed.

Inventive Principle:

Principle #16Partial or excessive action

4Extent of automation

If automatic ID photograph creation system is used, then photograph processing is automated, but the system is large and installation site is limited

Engineering Contradiction:

Improvephotograph processing automationVSAvoidsystem size

Core Design Contradiction:

Extent of automationVSDevice complexity

Solution Approach 1:

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The detection system is designed to handle multiple types of target objects and variations using a unified framework that combines template matching, characteristic-based detection, and neural network methods. This multi-functionality reduces the need for separate specialized systems.

Inventive Principle:

Principle #6Universality (Multi-functionality)

Data Source

PatentUS7542591B2Target object detecting method, apparatus, and program

Publication Date: 2009.06.02 FUJIFILM CORP

AI summary

Detecting a predetermined target object from a digital image reliably and rapidly. The standard face detecting section performs detection process for detecting a standard face from a photograph image. The eyeglassed face detecting section performs detection process for detecting an eyeglassed face from the photograph image from which no face has been detected by the standard face detecting section. The whiskered face detecting section performs detection process for detecting a whiskered face from the photo image from which no face has been detected by the standard face detecting section and eyeglassed face detecting section.