Machine-Learned Imaging Settings for In-Focus Composite Images
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In capturing images of subjects at varying distances, existing methods require numerous images for depth composition, leading to increased processing load and difficulty in determining optimal settings, especially for inexperienced photographers.
Innovation Solution
A learning apparatus that generates imaging settings using machine learning based on teacher data, including subject and composite images, to determine suitable in-focus positions for generating composite images.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the number of captured images is increased to improve composite image resolution, then the perceived resolution of the composite image is improved, but the processing load increases
Solution Approach 1:
The patent applies parameter changes by using machine learning to automatically determine optimal imaging parameters (number of captured images, in-focus positions, shooting intervals) based on subject characteristics. This resolves the contradiction by finding the optimal balance point where sufficient resolution is achieved without unnecessary increase in processing load, as the learning apparatus identifies the minimum required parameters for quality composite images.
Solution Approach 2:
The learning apparatus enables self-service by automatically determining imaging settings without requiring user experience or manual trial-and-error. The system uses machine learning models to autonomously optimize the number of captured images and their parameters, eliminating the need for users to manually adjust settings while achieving high-resolution composite images with appropriate processing load.
2Measurement precision
If the number of captured images is increased to improve composite image resolution, then the perceived resolution of the composite image is improved, but the time required for image capture and processing increases
Solution Approach 1:
The patent optimizes time by using machine learning to determine the optimal number of captured images and their parameters. The learning apparatus calculates the minimum necessary images required for sufficient resolution, avoiding unnecessary captures. This resolves the time-resolution contradiction by identifying the optimal parameter set that achieves quality results with minimal capture and processing time.
Solution Approach 2:
The learning apparatus performs preliminary action by pre-determining optimal imaging parameters before actual image capture. The machine learning model predicts the ideal number of images and their settings based on subject characteristics, allowing the system to prepare and execute captures efficiently without trial-and-error adjustments during the shooting process.
3Measurement precision
If imaging settings are determined by trial and error based on photographer experience, then optimal settings can be obtained, but the ease of operation deteriorates for inexperienced photographers
Solution Approach 1:
The learning apparatus implements self-service by automatically determining optimal imaging settings without requiring user expertise. The machine learning model autonomously analyzes subject characteristics and generates appropriate parameters (number of images, in-focus positions, intervals), completely eliminating the need for user trial-and-error while maintaining optimal settings quality.
Solution Approach 2:
The learning apparatus acts as an intermediary between the user and the complex imaging parameters. Instead of requiring users to directly understand and adjust multiple technical parameters, the system serves as a mediator that translates simple user input (subject type) into optimized imaging settings, making the operation easy while preserving precision.
4Measurement precision
If the number of captured images is increased to improve composite image quality, then the perceived resolution is improved, but the quantity of data to be processed increases
Solution Approach 1:
The patent applies parameter changes by using machine learning to optimize the number of captured images based on subject characteristics. The learning apparatus identifies the minimum necessary quantity of images required to achieve sufficient resolution, preventing unnecessary data generation. This resolves the contradiction by finding the optimal parameter that balances quality output with manageable data quantity.
Data Source
AI summary
A learning apparatus includes a model generating unit configured to generate, by using teacher data including a subject and a composite image generated from a plurality of images captured based on settings predetermined for the subject, a model for generating imaging settings suitable for generating a composite image of the subject through machine learning. The plurality of images are images captured at in-focus positions different in an optical axis direction. The settings include at least the in-focus positions. The composite image is an image generated by extracting in-focus regions from the plurality of images.