Topological Coding for Visual Search

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current mobile visual search systems face high computational costs and communication expenses due to the large size of visual descriptors, which hinder real-time operations and compromise matching accuracy when searching large image repositories.

Innovation Solution

A visual search system generates a topologically encoded vector from a point set of an image, using a graph Laplacian matrix and affinity matrix, which is invariant to rotation and scaling, and can be compressed using methods like discrete cosine transform, facilitating efficient image search and identification.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If visual descriptors (feature points with locations) are used for image search, then matching accuracy is improved, but communication cost and data size increase significantly

Engineering Contradiction:
Improvematching accuracyVSAvoiddata size
Core Design Contradiction:
Measurement precisionVSQuantity of substance

Solution Approach 1:

The patent extracts only the essential topological relationships and geometric constraints from complete visual descriptors, representing images through a compact set of topological codes that capture the essential structure without transmitting full feature point data. This extraction reduces data size while preserving matching capability.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

Instead of transmitting actual feature point coordinates and descriptors, the patent creates a simplified topological copy that represents the essential spatial relationships. This topological code serves as a compact representation that can be used for image search without requiring the original detailed descriptor data.

Inventive Principle:
Principle #26Copying

2Measurement precision

If complete images are sent for visual search, then matching accuracy is maintained, but computational cost and communication expense become prohibitively high

Engineering Contradiction:
Improvematching accuracyVSAvoidreal-time operation capability
Core Design Contradiction:
Measurement precisionVSProductivity

Solution Approach 1:

The patent extracts essential topological and geometric information from complete images, creating a simplified representation that retains matching accuracy while enabling real-time processing. The topological code captures the essential structure needed for search operations.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The patent transforms image data from high-dimensional pixel values to a compact topological parameter representation. This parameter transformation reduces computational complexity while preserving the essential features needed for accurate image matching and search operations.

Inventive Principle:
Principle #35Parameter changes

3Quantity of substance

If feature point size is reduced to lower communication cost, then data transmission efficiency is improved, but searching performance and matching accuracy are compromised

Engineering Contradiction:
Improvefeature point sizeVSAvoidsearching performance
Core Design Contradiction:
Quantity of substanceVSMeasurement precision

Solution Approach 1:

The patent changes the parameter representation from detailed feature point coordinates to topological codes that encode spatial relationships. This parameter transformation maintains searching performance by preserving essential geometric constraints while significantly reducing feature point size for efficient transmission.

Inventive Principle:
Principle #35Parameter changes

4Manufacturing precision

If high-resolution image data is transmitted, then image quality is maintained, but communication cost and processing time increase

Engineering Contradiction:
Improveimage qualityVSAvoidprocessing time
Core Design Contradiction:
Manufacturing precisionVSLoss of time

Solution Approach 1:

The patent extracts essential topological and geometric information from high-resolution images, creating a compact representation that maintains image quality for search purposes while dramatically reducing processing time. The topological code preserves the essential structure needed for accurate matching without requiring full high-resolution data processing.

Inventive Principle:
Principle #2Taking out (Extraction)

Data Source

PatentEP2712453B1Image topological coding for visual search
Publication Date: 2020.04.15 HUAWEI TECH CO LTD
  • EP2712453B1 patent drawingFigure 1(a)~1(b)
  • EP2712453B1 patent drawingFigure 2
  • EP2712453B1 patent drawingFigure 3

AI summary

A method and an apparatus for processing images generate a first vector of a first number dimension for the image from a first number of points of the image based on topological information of the first number of points, and the first vector for the image is invariant to rotation and scaling in creating the image. The first number of points can be locations of a set of rotation and scaling invariant feature points of the image, and the generated first vector can be a graph spectrum of a pair-wise distance matrix generated from the first number of points of the image.