Adaptive Image Pyramid Processing for Scale Level Selection

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing image pyramid techniques for feature point extraction and image processing are inefficient due to prolonged operation times and increased false positive rates, as they require repeated processing across multiple scale levels.

Innovation Solution

An adaptive image processing apparatus and method that selectively identifies a representative scale level for image processing within an image pyramid, reducing the need for processing across all scale levels by using a first matching unit to select candidate images, a second matching unit to determine a suitable scale level, and an image processing unit to perform operations based on this scale level.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If image processing is performed on all images having various scale levels in the image pyramid, then comprehensive feature point detection is achieved, but operation time increases significantly

Engineering Contradiction:

Improvefeature point detection accuracyVSAvoidoperation time

Core Design Contradiction:

ReliabilityVSLoss of time

Solution Approach 1:

The patent segments the image pyramid processing into distinct levels: first performing processing on a downsampled low-resolution image pyramid to identify candidate regions, then selectively processing only those regions in the full-resolution image pyramid. This segmentation avoids processing all images at full resolution, thereby reducing operation time while maintaining detection accuracy.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent applies partial action by performing comprehensive image processing only on selected scale levels and candidate regions rather than on all images in the image pyramid. The downsampled image pyramid processing serves as a preliminary filter, applying processing partially to reduce the set of images requiring full-resolution processing.

Inventive Principle:

Principle #16Partial or excessive action

2Measurement precision

If feature point detection is performed on entire area of each image at multiple scale levels, then scale-invariant feature extraction is achieved, but false positive rate increases

Engineering Contradiction:

Improvescale-invariant feature extraction accuracyVSAvoidfalse positive rate

Core Design Contradiction:

Measurement precisionVSObject-generated harmful factors

Solution Approach 1:

The patent performs preliminary processing on downsampled images to identify candidate regions before processing the full-resolution images. This preliminary action filters out regions unlikely to contain relevant features, so that subsequent processing on full-resolution images is applied only to promising candidates, reducing false positives while maintaining scale-invariant detection capability.

Inventive Principle:

Principle #10Preliminary action

3Reliability

If repeated image processing is performed across multiple scale levels, then comprehensive image analysis is achieved, but processing speed decreases

Engineering Contradiction:

Improveimage analysis completenessVSAvoidprocessing speed

Core Design Contradiction:

ReliabilityVSProductivity

Solution Approach 1:

The patent segments the processing hierarchy into two levels: downsampled image pyramid processing for broad candidate identification, and full-resolution processing for detailed analysis of selected candidates. This segmentation enables comprehensive analysis at appropriate resolutions while improving processing speed by avoiding repeated full-resolution processing across all scale levels.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent applies partial processing by performing detailed image processing only on candidate regions identified from downsampled images, rather than repeatedly processing all images at full resolution across multiple scale levels. This selective approach maintains analysis completeness for relevant regions while significantly improving processing speed.

Inventive Principle:

Principle #16Partial or excessive action

Data Source

PatentUS9189867B2Adaptive image processing apparatus and method based in image pyramid

Publication Date: 2015.11.17 HANWHA VISION CO LTD

AI summary

An adaptive image processing apparatus includes a first matching unit selecting representative candidate images from among a plurality of images included in an image pyramid and calculating a first matching score between each of the representative candidate images and a target model; a second matching unit selecting one scale level from among scale levels of the representative candidate images based on the first matching score and calculating a second matching score between each of images having scale levels included in a dense scale level range with respect to the selected scale level and the target model; a representative scale level selecting unit selecting at least one of the scale levels included in the dense scale level range as a representative scale level based on the second matching scores; and an image processing unit performing image processing based on an image having the selected representative scale level.