Image Region Segmentation Using Luminance Normalization and Neighbor Graphs
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing techniques fail to accurately segment images with shadows as foreground regions, incorrectly classifying shadowed areas as background when region segmentation is performed under backlit or dark conditions.
Innovation Solution
An image processing apparatus that detects dark regions by normalizing luminance values and generates neighbor graphs to segment images into foreground and background regions, using a graph cut method to integrate adjacent fundamental regions based on edge connectivity and luminance differences.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If region segmentation is performed using conventional techniques on images with shadows, then the segmentation process can be completed, but shadow regions are incorrectly classified as background regions
Solution Approach 1:
The patent performs preliminary detection of shadow regions by comparing luminance values of fundamental regions against threshold values before performing region segmentation. By identifying and marking shadow regions in advance, the system prevents these regions from being incorrectly classified as background during the subsequent segmentation process, thereby resolving the contradiction between completing segmentation and maintaining classification precision.
Solution Approach 2:
The patent applies different processing strategies to different regions based on their luminance characteristics. Shadow regions with low luminance values are identified and handled differently from normal foreground regions. By assigning special treatment to shadow regions (marking them as potential foreground while adjusting segmentation criteria), the system maintains high classification precision for all region types without compromising overall segmentation reliability.
2Measurement precision
If fundamental regions with low luminance values are treated as shadow regions, then shadow detection is achieved, but adjacent skin regions may be incorrectly separated
Solution Approach 1:
The patent dynamically adjusts segmentation criteria based on the detected luminance characteristics of fundamental regions. When a fundamental region has a luminance value below a threshold, it is marked as a shadow region and special processing is applied. The system then adjusts the edge weightings and integration criteria in the neighbor graph to ensure that shadow regions are properly integrated with adjacent skin regions, preventing incorrect separation while maintaining accurate shadow detection.
Solution Approach 2:
The patent introduces an intermediary processing step that involves creating a neighbor graph with adjusted edge weights. This intermediary structure allows the system to consider both luminance differences and spatial relationships when determining region integration. The neighbor graph acts as a mediator that balances the tendency to separate low-luminance shadow regions with the tendency to integrate adjacent skin regions, resolving the contradiction between detection precision and integration reliability.
Data Source
AI summary
An image processing apparatus for region segmentation of an obtained image is provided. The image processing apparatus may include a detecting unit detecting, among plural fundamental regions set to an image, a fundamental region satisfying a predetermined condition as an appropriate region. The image processing apparatus may further include a specifying unit specifying adjacent states of fundamental regions excluding the appropriate region among the plural fundamental regions. The image processing apparatus may also include a region segmentation unit performing region segmentation on each component included in the image based on the adjacent states of the fundamental regions excluding the appropriate region specified by the specifying unit.