Compressed Histogram Face Descriptors for Bandwidth Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current face image analysis methods require high computation capabilities and suffer from limited face detection performance due to the large size of face descriptors, which hinders transmission, storage, and network bandwidth efficiency.
Innovation Solution
The method involves determining histogram-based face descriptors for face image regions and compressing them to reduce their size by up to 50-fold, maintaining effectiveness for face detection and analysis, thereby reducing memory footprint, storage latency, and network bandwidth consumption.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional face descriptors are used for face image analysis, then face detection performance is maintained, but the size of face descriptors becomes large which increases memory footprint and storage latency
Solution Approach 1:
The patent extracts only the essential information from the original face descriptors by identifying and retaining only the most significant histogram bins that contribute to face recognition accuracy. This extraction process reduces the descriptor size while maintaining recognition performance, directly addressing the contradiction between descriptor size and detection performance.
Solution Approach 2:
The patent changes the parameter representation by transforming face descriptors from their original high-dimensional form into a compressed representation using histogram-based features with reduced precision. This parameter transformation achieves significant size reduction (up to 50-fold) while preserving the essential characteristics needed for accurate face detection.
2Adaptability or versatility
If traditional face descriptors are transmitted over network, then face analysis can be performed remotely, but network bandwidth consumption increases
Solution Approach 1:
The patent extracts only the necessary face descriptor information for remote analysis, removing redundant data while preserving essential recognition features. This extraction enables efficient network transmission by sending only the compressed essential information rather than complete face images or full-resolution descriptors.
Solution Approach 2:
The patent transforms face descriptors into a compressed parameter representation that is optimized for network transmission. This parameter change reduces the data volume significantly, allowing remote face analysis to be performed with minimal bandwidth consumption while maintaining analysis accuracy.
3Productivity
If face descriptors are stored in memory, then face analysis operations can be performed, but storage access latency increases due to large descriptor size
Solution Approach 1:
The patent extracts only the essential face descriptor components needed for analysis operations, removing unnecessary data that would increase storage access time. This extraction enables faster memory access by reducing the amount of data that needs to be read and processed during face analysis operations.
Solution Approach 2:
The patent changes the storage representation of face descriptors into a compressed format that reduces memory footprint and access latency. This parameter transformation allows the system to maintain face analysis capability while significantly reducing the time required for storage access operations.
Data Source
AI summary
Methods and apparatuses are provided for facilitating face image analysis. A method may include determining a histogram-based face descriptor for each of a plurality of regions of a face image. The method may further include compressing the histogram-based face descriptors to generate a plurality of compressed face descriptors describing the plurality of regions of the face image. Corresponding apparatuses are also provided.