Multi-View Face Detection Using Boosting Chain Hierarchy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional face detection methods face challenges in handling variations in facial appearances, lighting, and pose, leading to complex and nonlinear face data distributions, resulting in inefficient detection processes with high computation costs and low precision.
Innovation Solution
A multiple-step face detection algorithm employing linear-filtering, a boosting chain hierarchy structure, and post-filtering processes, including SVM and color-filtering, to enhance detection accuracy and efficiency, combined with a two-level hierarchy in-plane pose estimator for handling multi-view face detection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a view-based method with multiple face models is used to handle pose variations, then detection accuracy for multi-view faces is improved, but computation complexity increases significantly
Solution Approach 1:
The patent segments the face detection task into multiple independent view channels (frontal, left profile, right profile, etc.), with each channel having its own dedicated detector. This segmentation allows the system to handle different pose variations separately, improving detection accuracy for multi-view faces while keeping each individual detector relatively simple, thus managing computation complexity through modular architecture.
Solution Approach 2:
The patent introduces a view-based dimension to the detection problem, organizing detectors not just by spatial location but by pose angle. By transforming the detection problem into a multi-dimensional space encompassing both spatial and angular dimensions, the system can efficiently handle pose variations through a structured approach that balances accuracy and computational requirements.
2Productivity
If a cascade of boosting classifiers is used for fast frontal face detection, then detection speed is improved, but detection precision deteriorates due to discarded inter-layer information
Solution Approach 1:
The patent merges the cascade structure with a feedback mechanism that preserves and utilizes inter-layer information. Instead of discarding information from previous boosting classifier layers, the system combines their outputs with the current layer's predictions, creating a hybrid approach that maintains the speed advantage of cascades while recovering the precision loss through information reuse and integration across layers.
Solution Approach 2:
The patent introduces feedback loops that allow the system to learn from previous detection layers and adjust subsequent classifications accordingly. This feedback mechanism enables the boosting cascade to refine its predictions by incorporating information from earlier stages, thereby improving detection precision while preserving the fast processing capability of the cascade architecture.
3Reliability
If bootstrap procedures are used to collect non-face samples during training, then classifier training is improved, but the complexity of classification increases in later stages
Solution Approach 1:
The patent applies partial bootstrap procedures, selectively applying them only to certain stages or subsets of the training process rather than uniformly across all stages. This partial application allows the system to benefit from the reliability improvements of bootstrap sampling in early training stages while avoiding the excessive complexity accumulation that would occur if bootstrap procedures were applied throughout the entire training process, especially in later stages.
Data Source
AI summary
Face detection techniques are provided that use a multiple-stage face detection algorithm. An exemplary three-stage algorithm includes a first stage that applies linear-filtering to enhance detection performance by removing many non-face-like portions within an image, a second stage that uses a boosting chain that is adopted to combine boosting classifiers within a hierarchy “chain” structure, and a third stage that performs post-filtering using image pre-processing, SVM-filtering and color-filtering to refine the final face detection prediction. In certain further implementations, the face detection techniques include a two-level hierarchy in-plane pose estimator to provide a rapid multi-view face detector that further improves the accuracy and robustness of face detection.