Local Feature Descriptor Extraction via Importance-Based Point Selection

Solution Overview

Problem

Existing local feature descriptor extraction methods generate large feature descriptors that increase communication and processing times, and reduce memory capacity, making them unsuitable for large-scale image searches, and also limit targets to feature points with high reproducibility, leading to similar issues.

Innovation Solution

A local feature descriptor extracting apparatus that detects feature points, selects a predetermined number of points in order of decreasing importance, and generates local feature descriptors for these selected points, reducing descriptor size while maintaining accuracy through a feature point detecting unit, a feature point selecting unit, and a local feature descriptor generating unit.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If local feature descriptors are generated for all detected feature points, then the completeness of feature representation is improved, but the descriptor size increases, leading to increased communication time, processing time, and memory usage

Engineering Contradiction:

Improvecompleteness of feature representationVSAvoidcommunication time and processing time

Core Design Contradiction:

ReliabilityVSLoss of time

Solution Approach 1:

The patent extracts only the most important feature points from the set of all detected feature points. The importance evaluation unit assesses each feature point's significance, and the selection unit extracts only those feature points that meet the importance threshold, thereby generating descriptors for a subset rather than all feature points, which reduces descriptor size while maintaining representation quality

Inventive Principle:

Principle #2Taking out (Extraction)

Solution Approach 2:

The patent changes the parameter of feature point selection from including all feature points to including only high-importance feature points. By introducing an importance evaluation mechanism that assigns weights or scores to feature points based on their discriminative power and reliability, the system transforms the descriptor generation process to operate on a filtered subset, thus reducing communication and processing time

Inventive Principle:

Principle #35Parameter changes

2Reliability

If local feature descriptors are generated for all detected feature points, then the completeness of feature representation is improved, but the memory capacity required increases, reducing the number of images that can be stored

Engineering Contradiction:

Improvecompleteness of feature representationVSAvoidmemory capacity

Core Design Contradiction:

ReliabilityVSQuantity of substance

Solution Approach 1:

The patent extracts only the most important feature points from the set of all detected feature points. The importance evaluation unit assesses each feature point's significance, and the selection unit extracts only those feature points that meet the importance threshold, thereby generating descriptors for a subset rather than all feature points, which reduces descriptor size while maintaining representation quality

Inventive Principle:

Principle #2Taking out (Extraction)

Solution Approach 2:

The patent changes the parameter of feature point selection from including all feature points to including only high-importance feature points. By introducing an importance evaluation mechanism that assigns weights or scores to feature points based on their discriminative power and reliability, the system transforms the descriptor generation process to operate on a filtered subset, thus reducing communication and processing time

Inventive Principle:

Principle #35Parameter changes

3Reliability

If feature points with high reproducibility are selected as targets, then the reliability of feature matching is improved, but the number of feature points is limited, leading to increased processing time and reduced memory efficiency

Engineering Contradiction:

Improveaccuracy of feature matchingVSAvoidprocessing time

Core Design Contradiction:

ReliabilityVSLoss of time

Solution Approach 1:

The patent changes the selection criterion from requiring high reproducibility to using importance evaluation that considers multiple factors including but not limited to reproducibility. This allows the system to select feature points that are most discriminative for object identification while maintaining matching reliability, thereby improving processing efficiency

Inventive Principle:

Principle #35Parameter changes

Solution Approach 2:

The patent applies different evaluation criteria to different feature points based on their local characteristics. Rather than uniformly requiring high reproducibility for all feature points, the system evaluates each feature point's importance in the context of its local image region, allowing selective processing that maintains accuracy while reducing overall processing time

Inventive Principle:

Principle #3Local quality

Data Source

PatentUS9805282B2Local feature descriptor extracting apparatus, method for extracting local feature descriptor, and program

Publication Date: 2017.10.31 NEC CORP

AI summary

The size of a feature descriptor is reduced, with the accuracy of object identification maintained. A local feature descriptor extracting apparatus includes a feature point detecting unit configured to detect a plurality of feature points in an image and output feature point information that is information on each of the feature points, a feature point selecting unit configured to select the predetermined number of feature points from the detected plurality of feature points in order of decreasing importance based on the feature point information, and a local feature descriptor generating unit configured to generate a local feature descriptor that is a feature descriptor in a local region for each of the selected feature points.