Image Retrieval Device Dynamic Feature Selection

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing image retrieval methods using local feature quantities are inefficient due to high calculation and matching times, particularly when detecting specific patterns with varying scales and rotations, as they require extensive computation for scale-invariant and rotation-invariant features.

Innovation Solution

The system optimizes detection time by selectively using rotation-invariant local feature quantities or scale-invariant and rotation-invariant local feature quantities based on user-defined scale correspondence conditions, reducing computation by generating parasitic patterns for scale correspondence and employing methods like SIFT for feature extraction.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If scale-invariant and rotation-invariant local feature quantities are used to detect patterns with varying scales and rotations, then detection accuracy is improved, but calculation time and matching time increase significantly

Engineering Contradiction:

Improvedetection accuracyVSAvoidcalculation time

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The patent dynamically selects between rotation-invariant and scale-invariant local feature quantities based on user-defined retrieval conditions. When scale correspondence is not required, only rotation-invariant features are used. When scale correspondence is required, scale-invariant features are selected. This dynamic adaptation optimizes the balance between detection accuracy and calculation time based on actual needs.

Inventive Principle:

Principle #15Dynamics

Solution Approach 2:

The patent changes the parameter of local feature quantity type (rotation-invariant vs. scale-invariant) based on retrieval conditions. By adjusting this parameter according to whether scale correspondence is needed, the system achieves optimal performance for each specific case without always incurring the high computational cost of scale-invariant features.

Inventive Principle:

Principle #35Parameter changes

2Loss of time

If rotation-invariant local feature quantities are used without scale invariance, then calculation time is reduced, but detection accuracy for patterns with varying scales deteriorates

Engineering Contradiction:

Improvecalculation timeVSAvoiddetection accuracy

Core Design Contradiction:

Loss of timeVSMeasurement precision

Solution Approach 1:

The system dynamically switches between rotation-invariant and scale-invariant local feature quantities based on whether scale correspondence is required in the retrieval conditions. This ensures that scale information is only processed when needed, maintaining accuracy for scale variations while avoiding unnecessary computational overhead.

Inventive Principle:

Principle #15Dynamics

Solution Approach 2:

The patent changes the invariance parameters of local feature quantities according to retrieval conditions. When scale correspondence is not required, only rotation invariance is applied. When scale correspondence is required, full scale and rotation invariance is applied, optimizing the balance between computational efficiency and detection accuracy.

Inventive Principle:

Principle #35Parameter changes

3Reliability

If the number of local feature quantity pairs to be matched is increased to improve detection completeness, then matching time increases proportionally

Engineering Contradiction:

Improvedetection completenessVSAvoidmatching time

Core Design Contradiction:

ReliabilityVSLoss of time

Solution Approach 1:

The patent extracts and uses only the necessary local feature quantities based on retrieval conditions. By taking out only the relevant features (rotation-invariant or scale-invariant as needed) rather than processing all possible feature pairs, the system maintains detection completeness while significantly reducing matching time.

Inventive Principle:

Principle #2Taking out (Extraction)

4Adaptability or versatility

If multiple types of local feature quantities are calculated to cover different retrieval conditions, then adaptability is improved, but device complexity increases

Engineering Contradiction:

Improveretrieval condition adaptabilityVSAvoidsystem complexity

Core Design Contradiction:

Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The patent implements a dynamic selection mechanism that chooses between rotation-invariant and scale-invariant local feature quantities based on retrieval conditions. This dynamic approach provides adaptability to different retrieval scenarios without requiring the system to simultaneously maintain multiple complex feature calculation pipelines, thus avoiding excessive device complexity.

Inventive Principle:

Principle #15Dynamics

Data Source

PatentUS10360471B2Image retrieving device, image retrieving method, and recording medium

Publication Date: 2019.07.23 CANON KK

AI summary

An image retrieving device capable of retrieving an image using a local feature quantity, the image retrieving device includes a memory, and a processor in communication with the memory, the processor being configured to perform operations including receiving a designation of a retrieval condition corresponding to a retrieval source image, determining whether to use a feature quantity of a local image having scale invariance on the basis of the retrieval condition, and performing control to calculate the feature quantity of the local image having scale invariance if it is determined to use the feature quantity of the local image having scale invariance, and performing control not to calculate the feature quantity of the local image having scale invariance if it is determined not to use the feature quantity of the local image having scale invariance.