Intraluminal Image Dark Portion Extraction via Gradient Vector Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Intraluminal images acquired by medical observation devices, such as endoscopes, often include dark portion areas that are unnecessary for observation and diagnosis, making it challenging to detect abnormal portions like mucous membranes or bleeding areas due to the presence of areas with low luminance that are similar to dark portions.
Innovation Solution
An image processing apparatus and method that detects low-luminance areas in intraluminal images, calculates peripheral feature data, and determines whether these areas are dark portion areas based on the concentration ratio of gradient vectors, allowing for accurate extraction of dark portion areas and subsequent focus on relevant abnormal areas.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If dark portion areas are extracted using simple luminance thresholding, then processing speed is improved, but measurement precision of dark portion area boundaries deteriorates
Solution Approach 1:
The patent segments the dark portion extraction process into two distinct phases: (1) initial rough extraction using luminance thresholding to quickly identify candidate regions, and (2) refined boundary determination using gradient vector analysis to precisely define the dark portion boundaries. This segmentation allows each method to be optimized for its specific purpose without compromising overall performance.
Solution Approach 2:
The patent performs preliminary action by first extracting candidate dark portion areas using simple luminance thresholding before applying the more computationally intensive gradient vector analysis. This preliminary extraction narrows down the processing scope, allowing the precise boundary detection to be applied only where needed, thus maintaining processing efficiency while improving accuracy.
2Measurement precision
If gradient vector analysis is applied to all low-luminance areas, then measurement precision of dark portion areas is improved, but device complexity increases
Solution Approach 1:
The patent applies local quality by using different processing methods for different regions: simple luminance thresholding is applied to the entire image for initial extraction, while gradient vector analysis is applied locally only to the boundaries of identified dark portion candidates. This localized application of complex processing reduces overall computational complexity while maintaining high precision where it matters most.
Solution Approach 2:
The patent implements partial action by applying the computationally intensive gradient vector analysis only to specific regions (dark portion candidates) rather than to the entire image. This selective application performs the precise analysis only where necessary, reducing overall processing complexity while still achieving high measurement precision for dark portion areas.
3Reliability
If color information is used for pixel exclusion, then reliability of mucous membrane detection is improved, but loss of information about luminance distribution increases
Solution Approach 1:
The patent segments the information processing into two independent streams: color information processing for reliable mucous membrane detection and luminance information processing for dark portion extraction. By maintaining these as separate processing channels, the patent prevents loss of luminance distribution information while still utilizing color information for its intended purpose of reliable tissue detection.
Solution Approach 2:
The patent introduces an intermediary approach by using gradient vectors as a mediator that operates on luminance information independently of color-based pixel exclusion. The gradient vector analysis captures luminance distribution characteristics without being influenced by color-based filtering, thus preserving luminance information that would otherwise be lost when color thresholds are applied.
Data Source
AI summary
An image processing apparatus includes a low-luminance area detecting unit that detects a low-luminance area based on pixel values of pixels of the intraluminal image; a peripheral feature data calculating unit that calculates peripheral feature data based on pixel values of periphery of the low-luminance area; and a dark portion area determining unit that determines whether or not the low-luminance area is the dark portion area based on the peripheral feature data.