Image Search System Using Cascade Feature Extraction

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current image search engines face challenges in accurately retrieving similar images from external databases without pre-defined descriptive information, leading to poor search results and user inconvenience, as they rely heavily on user-inputted category and style information which may be inaccurate.

Innovation Solution

The system automatically determines category and descriptive information for query images by comparing visual features with those in the database, using a cascade-type re-search method that extracts and compares global and local features, eliminating the need for user-provided metadata and improving search accuracy.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If the system requests users to provide category, style information, and other descriptive information when inputting query images, then search accuracy is improved, but the search process becomes cumbersome and user experience deteriorates

Engineering Contradiction:
Improvesearch accuracyVSAvoidsearch process convenience
Core Design Contradiction:
Measurement precisionVSEase of operation

Solution Approach 1:

The system performs self-service by automatically extracting category information, style information, and descriptive information from the query image itself using image recognition technology. This eliminates the need for users to manually input these details, thereby maintaining high search accuracy while significantly improving search process convenience and user experience.

Inventive Principle:
Principle #25Self-service

2Productivity

If users manually input descriptive information for query images, then search results may be obtained, but the inputted information may be inaccurate leading to incorrect search results

Engineering Contradiction:
Improvesearch efficiencyVSAvoidinformation accuracy
Core Design Contradiction:
ProductivityVSReliability

Solution Approach 1:

The patent replaces the manual mechanical input system with an automated image recognition system. Instead of relying on users to manually input descriptive information (which may be inaccurate), the system uses computer vision algorithms to automatically extract and analyze visual features from the query image, thereby improving both search efficiency and information accuracy simultaneously.

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

3Measurement precision

If the system conducts comprehensive visual feature extraction and comparison, then search accuracy is improved, but system resource consumption increases

Engineering Contradiction:
Improvesearch accuracyVSAvoidsystem resource consumption
Core Design Contradiction:
Measurement precisionVSUse of energy by moving object

Solution Approach 1:

The patent applies segmentation by dividing the visual feature extraction and comparison process into distinct modules: global feature extraction, local feature extraction, and hierarchical comparison. This modular approach allows the system to process images efficiently by extracting only relevant features at appropriate levels of detail, thereby maintaining high search accuracy while reducing overall system resource consumption compared to exhaustive full-image processing.

Inventive Principle:
Principle #1Segmentation

Data Source

PatentEP3028184B1Method and system for searching images
Publication Date: 2018.12.05 ALIBABA GROUP HOLDING LTD
  • EP3028184B1 patent drawingFigure 1A
  • EP3028184B1 patent drawingFigure 1B
  • EP3028184B1 patent drawingFigure 1C

AI summary

Embodiments of the present application relate to a method for searching images, a system for searching images, and a computer program product for searching images. A method for searching images is provided. The method includes receiving an input query image, extracting visual features from the inputted query image; determining a similarity of the visual features of the query image and visual features of images in an image database; determining category information, descriptive information, or a combination thereof associated with the query image based on category information, descriptive information, or a combination thereof of business objects corresponding to images having a similarity to the query image that complies with a precondition; conducting searches of the images based on the query image and the category information, the descriptive information, or a combination thereof associated with the query image; and returning search results.