Lesion Candidate Image Resizing With Aspect-Ratio-Preserving Padding

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing image processing techniques for endoscopic images reduce the diagnostic accuracy of lesion candidates by altering their two-dimensional shape and surface pattern.

Innovation Solution

An image processing device that includes resize and padding operations to maintain the aspect ratio and adjust the size of lesion candidate images, enhancing diagnostic accuracy by preserving the two-dimensional shape and surface pattern.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Illumination intensity

If image processing techniques (color interpolation, color separation, color balance adjustment, gamma correction, image enhancement processing) are applied to endoscopic images, then image quality and display characteristics are improved, but the diagnostic accuracy of lesion candidates is reduced

Engineering Contradiction:

Improveimage qualityVSAvoiddiagnostic accuracy

Core Design Contradiction:

Illumination intensityVSMeasurement precision

Solution Approach 1:

The patent segments the image processing workflow into two distinct pathways: a diagnostic pathway that preserves original image characteristics for accurate lesion detection, and a display pathway that applies enhancement processing for improved visual presentation. This segmentation allows both image quality improvement and diagnostic accuracy maintenance by directing different processing treatments to different usage scenarios.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent extracts the essential diagnostic information from enhanced images by comparing processed images with original images, identifying and isolating lesion candidate regions that maintain their characteristic features. This extraction process ensures that while display images receive enhancement processing, the diagnostic integrity of lesion characteristics is preserved through direct reference to unprocessed original data.

Inventive Principle:

Principle #2Taking out (Extraction)

2Ease of manufacture

If processing to change the state of lesion characteristics (two-dimensional shape and surface pattern) is performed, then image enhancement is achieved, but the diagnostic accuracy of lesion candidates is lowered

Engineering Contradiction:

Improveimage enhancementVSAvoidlesion characteristic accuracy

Core Design Contradiction:

Ease of manufactureVSManufacturing precision

Solution Approach 1:

The patent performs preliminary detection of lesion candidates using original, unprocessed images before any enhancement processing is applied. By identifying lesion regions in the original image first, the system ensures that subsequent enhancement processing on display images does not alter the fundamental characteristics of detected lesions, maintaining diagnostic accuracy while still providing enhanced visual presentation.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent creates and utilizes copies of original images for different purposes: one copy undergoes enhancement processing for display, while another copy remains unprocessed for diagnostic verification. This copying approach allows the system to provide enhanced images for viewing while maintaining accurate, unaltered references for diagnostic decision-making.

Inventive Principle:

Principle #26Copying

Data Source

PatentUS12541846B2Image processing device for preparing lesion candidate images for a machine learning model

Publication Date: 2026.02.03 NEC CORP

AI summary

In an image processing device, the image acquisition means acquires a lesion candidate image including a lesion candidate detected from an endoscopic image. The resize means performs resize processing for resizing a size of the lesion candidate image according to a target size while maintaining an aspect ratio of the lesion candidate image. The padding means performs padding processing for matching a size of a resized image obtained by the resize processing to the target size.