Image Clustering Using GIST Features and LSH Fingerprints
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image clustering methods are inefficient in identifying and grouping similar advertising images across different orders, leading to sparse distribution and reduced advertising delivery effectiveness.
Innovation Solution
The method employs Global Scene Semantic (GIST) feature extraction and Local Sensitive Hashing (LSH) algorithms to generate image fingerprints, calculating similarity between images and classifying them into categories based on a predetermined threshold, thereby facilitating fast clustering of similar advertising orders.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional image clustering methods are used to group similar advertising images, then the clustering process becomes computationally intensive and time-consuming, but the advertising orders remain sparsely distributed and delivery effectiveness is reduced
Solution Approach 1:
The patent segments the image analysis process into two distinct stages: first extracting GIST features to capture global scene semantics, then applying LSH algorithms to generate compact image fingerprints. This segmentation allows efficient similarity calculation without sacrificing clustering accuracy, as each stage focuses on specific aspects of image comparison.
Solution Approach 2:
The patent introduces image fingerprints as an intermediary representation between original images and clustering results. These fingerprints serve as compressed descriptors that preserve essential similarity information while enabling fast comparison. The LSH algorithm acts as a mediator that transforms high-dimensional GIST features into compact binary fingerprints suitable for efficient hashing and matching.
2Measurement precision
If comprehensive image analysis is performed to ensure accurate clustering, then the processing time and computational resources increase significantly
Solution Approach 1:
The patent performs preliminary action by extracting GIST features first to capture the global semantic content of images before applying LSH for fingerprint generation. This preliminary extraction of meaningful features ensures that subsequent similarity measurements are based on semantically relevant information, improving accuracy while avoiding redundant computations on raw pixel data.
Solution Approach 2:
The patent extracts only the essential GIST features that capture global scene semantics, discarding redundant detailed information. This extraction principle focuses computational resources on the most discriminative features for advertising image comparison, achieving accurate clustering with reduced processing requirements.
Data Source
AI summary
A method and a clustering system for image clustering, and a computer-readable storage medium are provided. The method includes: extracting a GIST feature of a first image and a GIST feature of a second image; obtaining an image fingerprint of the first image, based on the GIST feature of the first image and in conjunction with an LSH algorithm and obtaining an image fingerprint of the second image, based on the GIST feature of the second image and in conjunction with the LSH algorithm; calculating a similarity between the first and second images, based on the image fingerprints of the first and second images; and classifying the first image and the second image as a same category of image in a case that the similarity between the first image and the second image is larger than a predetermined similarity threshold.