Image Change Amount Extraction for Capsule Endoscopy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for detecting scene changes in continuous image sequences are inefficient, particularly in large datasets like those from capsule endoscopes, where many similar images are redundant, making it impractical for human observation.
Innovation Solution
An image processing apparatus and method that calculates a predetermined image change amount between images, adds this information to each image, and extracts a preset number of images based on this change, allowing for efficient detection and extraction of scene changes by prioritizing images with significant feature value changes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional threshold processing is used to detect scene changes in continuous image sequences, then scene change detection can be performed, but the number of images to be displayed becomes too large making it impractical for human observation
Solution Approach 1:
The patent extracts only the essential information needed for scene change detection by calculating image change amounts and extracting only the top N images with significant changes. This removes redundant images while preserving detection accuracy, directly addressing the contradiction between comprehensive detection and manageable quantity.
Solution Approach 2:
The patent changes the parameter used for image selection from simple threshold-based binary classification to a ranked extraction based on image change amounts. By sorting images according to their change amounts and selecting only the top N, the system reduces the quantity of images while maintaining detection reliability through parameter-based prioritization.
2Loss of information
If all images in the sequence are displayed to ensure complete information, then no information is lost, but the time required for observation becomes excessively long
Solution Approach 1:
The patent extracts only the most informative images (top N with highest change amounts) from the complete sequence. This extraction preserves critical scene change information while eliminating redundant images that would consume unnecessary observation time, resolving the contradiction between information completeness and time efficiency.
Solution Approach 2:
The patent performs preliminary calculation of image change amounts for all images in the sequence before extraction. This preliminary action ranks all images by importance, allowing the system to pre-identify which images are essential for observation, thereby reducing observation time without losing critical information.
3Reliability
If images are extracted based on time difference between scenes, then scene changes can be identified, but images with similar appearance patterns are incorrectly identified as scene changes
Solution Approach 1:
The patent changes the detection parameter from time-based difference to image content-based change amount calculation. By measuring actual image differences (using methods like sum of absolute differences or correlation coefficients) and ranking based on these quantitative measures, the system achieves more precise measurement that distinguishes true scene changes from similar appearance patterns.
Data Source
AI summary
An image processing apparatus for extracting images from a continuous image sequence includes a storage unit that stores image information about images constituting the image sequence; an image reading unit that reads the image information from the storage unit; and an image change amount calculating unit that calculates a predetermined image change amount between at least two images using the image information read by the image reading unit. The apparatus also includes an image change amount information adding unit that adds information about the image change amount calculated by the image change amount calculating unit to a corresponding image; and an image extracting unit that extracts a preset number of images from the image sequence based on the information added to each image by the image change amount information adding unit.