Image Selection Using Desirability Scores and Time Intervals
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional digital camera systems with high-speed continuous shooting capabilities require numerous operations to select desirable images with different contents, as existing methods either focus on temporally close images with high recognition scores or display images with low coincidence, leading to inefficiencies in selecting high-scoring, content-different images.
Innovation Solution
An image processing apparatus that calculates desirability scores using recognition scores and shooting time, sets a time interval, and selects images based on local maximum desirability scores and time intervals to efficiently identify and display best-shot images with distinct contents.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If images are selected in decreasing order of recognition scores without time interval constraints, then high recognition score images are prioritized, but only temporally close images are selected leading to redundant content
Solution Approach 1:
The patent divides the image selection process into multiple stages by introducing time interval thresholds. Images are first grouped by time intervals, then selected within each group. This segmentation prevents selecting multiple images with similar content within the same time period while still prioritizing high recognition score images across different time periods.
Solution Approach 2:
The patent adds a time dimension to the traditional recognition score-based selection. Instead of solely ranking images by recognition score, the system incorporates time interval as an additional constraint dimension, creating a two-dimensional selection criteria that balances score priority with temporal diversity.
2Adaptability or versatility
If images with low coincidence are displayed to avoid redundancy, then content diversity is improved, but images with low recognition scores become display targets reducing quality
Solution Approach 1:
The patent applies different selection criteria to different time intervals. Within each time interval group, images are selected based on their coincidence with previously selected images. This local quality approach ensures diversity within time periods while maintaining high recognition scores across the overall selection.
Solution Approach 2:
The system performs preliminary grouping of images by time intervals before applying the coincidence-based selection. This preliminary action organizes the image data structure to enable subsequent selective filtering that maintains both diversity and quality.
3Adaptability or versatility
If multiple operations are performed to find best-shots with different contents, then content variety is achieved, but the number of operations increases significantly
Solution Approach 1:
The patent implements continuous processing where images are evaluated and selected in a single pass through the time-interval grouped data structure. The system maintains a running list of selected images and continuously checks coincidence against this list, eliminating the need for multiple sequential operations to find diverse best-shots.
Data Source
AI summary
An image processing method includes calculating a desirability score which indicates a desirability of the image, for each of a series of images captured in a time-sequential manner, based on a recognition result of the image, and selecting a plurality of images as best-shots in which desirability scores are a local maximum value in a time-sequence, from the series of images.