Facial Feature Detection Using Head and Face Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing technologies face challenges in accurately detecting feature areas, such as face and pupil areas, due to reliance on single detection methods, which can lead to failure if initial partial area detection is incorrect, resulting in reduced processing speed and accuracy.
Innovation Solution
An image processing apparatus and method that employs multiple detection units for partial areas, including face, head, and pupil detection, allowing for alternative detection methods to ensure accurate feature area identification, even if one method fails, by using a combination of face detection units, head detection units, and pupil detection units to increase reliability and speed.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a single detection method is used to detect partial areas (face or head), then the device complexity is reduced, but the reliability of feature area detection deteriorates because the detection may fail under certain conditions
Solution Approach 1:
The detection system is segmented into multiple independent detection units: a face detection unit and a head detection unit. Each unit operates independently to detect partial areas using different detection methods, thereby improving overall detection reliability without requiring a single complex detection system
Solution Approach 2:
The detection apparatus is designed with multi-functionality by incorporating both face detection and head detection capabilities. This allows the system to adaptively select the appropriate detection method based on the imaging conditions, enhancing reliability while managing complexity through functional versatility
2Reliability
If multiple detection methods are employed to improve detection reliability, then the processing time increases due to performing multiple detection operations
Solution Approach 1:
The system performs preliminary detection using one method (e.g., face detection) first, then selectively performs additional detection using another method (e.g., head detection) only when needed. This staged approach ensures high reliability while minimizing unnecessary processing time by avoiding redundant detections
Solution Approach 2:
The detection system dynamically adjusts its operation by selectively activating different detection units based on imaging conditions. When the subject is close to the camera, head detection is prioritized; when farther away, face detection is used. This dynamic adaptation optimizes the balance between reliability and processing time
3Measurement precision
If the detection area is not properly limited before feature detection, then the detection accuracy may improve by searching the entire image, but the processing speed decreases due to larger search space
Solution Approach 1:
The image search space is segmented into two distinct partial areas: a face area detected by the face detection unit and a head area detected by the head detection unit. This segmentation allows subsequent feature detection (such as pupil detection) to be performed within smaller, more focused regions, thereby improving both accuracy and processing speed
Solution Approach 2:
The system performs preliminary detection of partial areas (face or head) before conducting feature detection. This preliminary action establishes appropriate detection boundaries that guide subsequent feature detection, ensuring high accuracy while maintaining processing speed by limiting the search space to relevant regions
Data Source
Figure 1
Figure 2
Figure 3
AI summary
An image processing apparatus (124), comprises first detection units (201, 202) for detecting a predetermined partial area from image data separately using a first method and a second method; and second detection units (203, 204) for detecting a predetermined feature area from data of the detected partial areas. The apparatus (124) selects, as a final detection result of the feature area, one of a detection result of the feature area for the partial area detected using the first method and a detection result of the feature area for the partial area detected using the second method, wherein the partial areas are related to an identical subject.