Shooting Control Using Existence Probability Distributions
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional shooting control methods often fail to accurately distinguish between subject and background areas, leading to inappropriate focus and exposure adjustments, as they do not effectively utilize the distribution of subject existence probabilities in captured image data.
Innovation Solution
An apparatus and method that detect subject areas within captured image data, utilizing existence probability distributions to prioritize data contributions based on position-specific probabilities, thereby enhancing shooting control, such as focus, white balance, and exposure adjustments.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If subject detection is performed using conventional methods, then a subject area can be detected, but the detected area includes both subject and background regions, leading to inaccurate shooting control
Solution Approach 1:
The subject area is segmented into multiple regions based on existence probability thresholds. High-probability regions (where the subject is likely to exist) are distinguished from low-probability regions (background areas). This segmentation allows shooting control to be applied differently to each region, improving accuracy without requiring complex manual intervention.
Solution Approach 2:
Different shooting control parameters are applied to different regions within the subject area. High-probability regions receive primary shooting control (e.g., focus adjustment), while low-probability regions receive secondary or no shooting control. This local differentiation ensures that shooting control is optimized for actual subject locations rather than treating the entire detected area uniformly.
2Reliability
If shooting control is performed based on the entire detected subject area, then coverage is maximized, but background areas are incorrectly included in focus and exposure adjustments
Solution Approach 1:
The subject area is divided into high-probability and low-probability regions based on pre-stored existence probability distributions. Shooting control is selectively applied to high-probability regions, ensuring that focus and exposure adjustments are made only where the subject is likely to be present. This segmentation improves reliability by excluding background areas from incorrect control adjustments.
Solution Approach 2:
Existence probability distributions are pre-calculated and stored for various subject types and shooting conditions. During actual shooting, the system retrieves these pre-computed distributions and applies them to segment the subject area, rather than performing complex real-time analysis. This preliminary preparation maintains processing efficiency while enabling accurate region differentiation.
3Ease of manufacture
If uniform shooting control is applied across the detected subject area, then processing is simplified, but the contribution of background areas degrades the quality of focus and exposure settings
Solution Approach 1:
The system assigns different weights to different regions within the subject area based on existence probabilities. High-probability regions are given higher weights, meaning their data contributes more to focus and exposure calculations. Low-probability regions (background areas) are given lower weights or excluded entirely. This weighted approach maintains implementation simplicity while significantly improving shooting control precision by reducing background interference.
Data Source
AI summary
Based on captured image data, a detection unit detects a subject area that partially includes a subject to be detected. A control unit performs shooting control based on data corresponding to the subject area among the captured image data, and on a distribution of existence probabilities of the subject in the subject area. The distribution is stored in a memory. The control unit performs the shooting control so that a contribution of data corresponding to a position with a first existence probability is larger than a contribution of data corresponding to a position with a second existence probability that is lower than the first existence probability.