Image Retrieval Using Global and Local Feature Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image retrieval methods face challenges in maintaining accuracy and efficiency, especially when dealing with large datasets and variations in viewpoint and illumination, as they often rely on either global feature extraction which is not robust or local feature extraction which lacks ideal retrieval accuracy.
Innovation Solution
A method that extracts both global and local features of images using a preset neural network model, followed by candidate image selection through feature matching and verification, optimizing the retrieval process by combining global and local features for improved accuracy and efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If image feature extraction is performed based on image feature descriptor, then retrieval accuracy is maintained under consistent viewpoint and illumination, but retrieval efficiency deteriorates when dealing with large number of images
Solution Approach 1:
The patent divides image features into two segments: global features extracted by CNN for initial candidate screening, and local features extracted by feature descriptors for final verification. This segmentation allows each method to operate on appropriate data scope, improving overall efficiency while maintaining accuracy.
Solution Approach 2:
The patent applies different feature extraction qualities to different parts of the retrieval process: global features provide coarse-grained quality for rapid filtering, while local features provide fine-grained quality for precise matching, optimizing the balance between speed and accuracy.
2Productivity
If convolutional neural network model is used for image retrieval, then retrieval efficiency is improved, but retrieval accuracy deteriorates
Solution Approach 1:
The patent merges two feature extraction approaches by combining global features from CNN with local features from feature descriptors. The CNN provides efficient global matching while feature descriptors add precise local verification, achieving both speed and accuracy.
Solution Approach 2:
The patent introduces candidate image sets as an intermediary stage between global feature matching and final local feature verification. This mediator structure allows the system to filter images in two stages, preventing direct comparison of all images while maintaining high accuracy through final local verification.
3Speed
If only global feature matching is performed, then retrieval speed is improved, but accuracy deteriorates under viewpoint and illumination variations
Solution Approach 1:
The patent performs preliminary global feature matching to quickly identify candidate images that may match the query. This preliminary action filters out obviously unrelated images before performing more computationally intensive local feature verification, maintaining speed while enabling accurate verification of promising candidates.
Data Source
AI summary
A method for retrieving an image is provided. The method includes: extracting a global feature and a local feature of an image to be retrieved, and a global feature and a local feature of an image to be recalled by employing a preset neural network model; determining a candidate image set by matching the global feature of the image to be retrieved with the global feature of the image to be recalled and matching the local feature of the image to be retrieved with the local feature of the image to be recalled; and determining a retrieval result from the candidate image set by performing local feature verification on the image to be retrieved and a candidate image in the candidate image set. An apparatus for retrieving an image, an electronic device, and a medium are further provided.