Face Detection Using Single Database and Image Transformation

Solution Overview

Problem

Existing face detection technologies require multiple face databases for detecting tilted or side faces, leading to complex training procedures and significant storage needs.

Innovation Solution

A face detection method and system that transforms original image blocks into multiple orientations using a single face database, employing parallel processing cores to determine if a face is present by comparing detecting result values against a threshold, thereby omitting the need for multiple databases and reducing storage requirements.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If multiple face databases are provided for detecting tilted or side faces, then the detection coverage is improved, but the storage space requirement increases and training procedures become complex

Engineering Contradiction:

Improvedetection coverageVSAvoidstorage space

Core Design Contradiction:

Adaptability or versatilityVSQuantity of substance

Solution Approach 1:

The patent segments the face detection task by dividing the original image into multiple transformed image blocks through geometric transformations (rotation, scaling, flipping). Each transformed block is then detected independently using a single face database, eliminating the need for multiple databases while maintaining comprehensive detection coverage for tilted and side faces.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent transforms the detection problem from a multi-database approach to a single-database approach by adding the dimension of image transformation. Instead of storing multiple databases for different face orientations, the system applies geometric transformations to the input image to generate multiple viewing angles, effectively solving the adaptability problem without increasing storage requirements.

Inventive Principle:

Principle #17Another dimension (Dimensionality change)

2Adaptability or versatility

If multiple face databases are provided for detecting tilted or side faces, then the detection coverage is improved, but the training procedures become complex

Engineering Contradiction:

Improvedetection coverageVSAvoidtraining procedures

Core Design Contradiction:

Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The patent segments the face detection task by dividing the original image into multiple transformed image blocks through geometric transformations (rotation, scaling, flipping). Each transformed block is then detected independently using a single face database, eliminating the need for multiple databases while maintaining comprehensive detection coverage for tilted and side faces.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent transforms the detection problem from a multi-database approach to a single-database approach by adding the dimension of image transformation. Instead of storing multiple databases for different face orientations, the system applies geometric transformations to the input image to generate multiple viewing angles, effectively solving the adaptability problem without increasing storage requirements.

Inventive Principle:

Principle #17Another dimension (Dimensionality change)

3Productivity

If multiple transformed image blocks are detected by parallel processing cores, then the processing efficiency is improved, but the device complexity increases

Engineering Contradiction:

Improveprocessing efficiencyVSAvoiddevice structure

Core Design Contradiction:

ProductivityVSDevice complexity

Solution Approach 1:

The patent segments the face detection task by dividing the original image into multiple transformed image blocks through geometric transformations (rotation, scaling, flipping). Each transformed block is then detected independently using a single face database, eliminating the need for multiple databases while maintaining comprehensive detection coverage for tilted and side faces.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent merges multiple detection results from parallel processing cores into a unified detection outcome. By combining the results from different transformed image blocks and using a maximum value comparison with a threshold, the system achieves efficient parallel processing while maintaining a relatively simple overall device structure.

Inventive Principle:

Principle #5Merging (Combining)

Data Source

PatentUS10282592B2Face detecting method and face detecting system

Publication Date: 2019.05.07 ICATCH TECH INC

AI summary

A face detecting method and a face detecting system are provided. The face detecting method includes the following steps: At least one original image block is received. The original image block is transformed by a transforming unit to obtain a plurality of different transformed image blocks. Whether each of the transformed image blocks contains a face is detected by a detecting unit according to only one identical face database and a detecting result value is outputted accordingly. The transformed image blocks are detected by a plurality of parallel processing cores. Whether a maximum of the detecting result values is larger than a threshold value is determined by a determiner. If the maximum of the detecting result values is larger than the threshold value, then the determiner deems that the original image block contains a face.