Human Body Key Point Feature Integration for Image Search
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing techniques face accuracy degradation when searching or classifying images with human bodies that have obscured parts or are in undesired poses or movements, as they rely on clear and fully detectable key points and desired poses or movements.
Innovation Solution
An image processing apparatus and method that detects key points of human body parts, computes feature values, and integrates them using user-specified methods to generate integrated feature values for improved search and classification accuracy, even when parts are obscured or in undesired poses, by complementing missing feature values from other images.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If an image with obscured human body parts or undesired poses is used for search or classification, then the availability of images increases, but accuracy is degraded
Solution Approach 1:
The patent merges feature values from multiple images showing the same human body part to create a comprehensive feature representation. By combining information from images with different poses and occlusion levels, the system achieves both broader image availability and maintained accuracy, as the integrated feature value captures the complete characteristics of the human body part across various viewing conditions
2Measurement precision
If multiple images are integrated to compensate for obscured parts, then accuracy is improved, but processing complexity increases
Solution Approach 1:
The patent segments the human body into multiple key points and processes feature extraction and integration separately for each key point. This segmentation approach simplifies the overall processing complexity by breaking down the complex task of integrating multiple images into manageable per-key-point operations, while still achieving improved accuracy through comprehensive feature integration
3Reliability
If feature values from multiple human bodies are integrated, then robustness against occlusions is improved, but computation time increases
Solution Approach 1:
The patent performs preliminary feature extraction and integration by pre-computing integrated feature values from multiple images before the actual search or classification operation. This preliminary action stores the integrated features for later use, reducing computation time during query operations while maintaining robustness against occlusions, as the comprehensive feature integration is already completed in advance
Data Source
AI summary
The present invention provides an image processing apparatus (100) including a skeleton structure detection unit (101) that executes processing of detecting a plurality of key points associated with each of a plurality of parts of a human body included in an image, a feature value computation unit (102) that computes a feature value of each of the key points being detected, an input unit (106) that receives a user input for specifying a method of integrating the feature values of the key points detected from each of a plurality of human bodies for each of the parts, and a processing unit (103) that computes an integrated feature value of each of the parts by performing integration for each of the parts by the method specified by the user input, and performs an image search or image classification, based on the integrated feature value.