Keypoint Selection for Image Analysis Data Reduction

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing image analysis methods require managing a large amount of data, particularly in mobile applications, due to the high number of keypoints and SIFT descriptors, leading to inefficiencies and increased computational load, which limits scalability and performance.

Innovation Solution

A method is proposed to reduce data processing by selecting an optimal subset of keypoints based on local feature relevance probabilities calculated using reference statistical distributions, and compressing descriptors and coordinates to minimize data transmission and processing demands.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If SIFT descriptors are used for accurate image comparison, then image recognition accuracy is improved, but the amount of data to be processed becomes excessive

Engineering Contradiction:
Improveimage recognition accuracyVSAvoidamount of data
Core Design Contradiction:
Measurement precisionVSQuantity of substance

Solution Approach 1:

The patent extracts and transmits only the most relevant keypoints and their descriptors from the image, rather than transmitting all keypoints. This is achieved by selecting keypoints based on criteria such as descriptor quality, location importance, and visual interest, thereby extracting only the essential data needed for accurate image comparison while reducing overall data volume.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The patent applies different processing qualities to different regions of the image by identifying and prioritizing keypoints in visually important areas (such as regions with high gradient magnitude, corners, or areas of interest) over less important regions. This allows the system to concentrate computational and transmission resources on locally significant features, maintaining accuracy where it matters most while reducing data from less critical areas.

Inventive Principle:
Principle #3Local quality

2Reliability

If all keypoints and descriptors are transmitted from mobile terminal to server, then complete image information is available for analysis, but data transmission efficiency deteriorates

Engineering Contradiction:
Improvecompleteness of image informationVSAvoiddata transmission efficiency
Core Design Contradiction:
ReliabilityVSProductivity

Solution Approach 1:

The mobile terminal extracts and transmits only the most relevant keypoints and their descriptors to the server, rather than transmitting complete image data or all keypoints. This extraction process filters out redundant information while preserving the essential features needed for accurate image analysis, thereby maintaining reliability while improving transmission efficiency.

Inventive Principle:
Principle #2Taking out (Extraction)

3Measurement precision

If a large number of keypoints are processed for accurate image comparison, then matching precision is improved, but computational load increases

Engineering Contradiction:
Improvekeypoint matching precisionVSAvoidcomputational load
Core Design Contradiction:
Measurement precisionVSPower

Solution Approach 1:

The system extracts and processes only the most relevant keypoints rather than all detected keypoints. By applying selection criteria based on descriptor quality, location importance, and visual interest, the system extracts a reduced subset of keypoints that maintains matching precision while significantly reducing the computational burden of processing and comparing all keypoints.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The system applies different processing intensities to different keypoints based on their local importance. Keypoints in visually significant regions (such as corners, high-gradient areas, or regions of interest) are processed with higher priority and detail, while keypoints in less important regions are processed with lower priority or excluded entirely. This local quality approach maintains precision for critical features while reducing overall computational load.

Inventive Principle:
Principle #3Local quality

Data Source

PatentEP2801056B1Image analysis
Publication Date: 2021.04.14 TELECOM ITALIA SPA
  • EP2801056B1 patent drawingFigure 1
  • EP2801056B1 patent drawingFigure 2A
  • EP2801056B1 patent drawingFigure 2B

AI summary

A method for processing an image is proposed. The method comprises identifying a first group of keypoints in the image. For each keypoint of the first group, the method provides for identifying at least one corresponding keypoint local feature related to said each keypoint; for said at least one keypoint local feature, calculating a corresponding local feature relevance probability; calculating a keypoint relevance probability based on the local feature relevance probabilities of said at least one local feature. The method further comprises selecting keypoints, among the keypoints of the first group, having the highest keypoint relevance probabilities to form a second group of keypoints, and exploiting the keypoints of the second group for analysing the image. The local feature relevance probability calculated for a local feature of a keypoint is obtained by comparing the value assumed by said local feature with a corresponding reference statistical distribution of values of said local feature.