Subject Distance Inference for Occlusion-Robust Camera Focusing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image capturing technologies struggle to accurately focus on a subject when there are blocking objects or low-contrast conditions, leading to significant errors in focus detection results, especially in low-illuminance environments or when using large f-numbers.
Innovation Solution
An inference apparatus using a machine learning model trained to suppress the contribution of distance information not based on the subject, by performing inference with multiple focus detection regions within a subject region, generating accurate distance information through a subject or range corresponding to the subject.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If statistical values are used to distinguish blocking objects in focus detection regions, then blocking objects can be identified, but focus detection accuracy deteriorates when the arm continuously blocks the face or in low-illuminance environments
Solution Approach 1:
The patent segments the subject region into multiple focus detection regions and processes distance information from each region independently. By dividing the detection task into smaller regional units, the system can identify blocking objects in specific regions without compromising focus detection accuracy in other regions. This segmentation allows selective application of blocking object detection only where needed.
Solution Approach 2:
The patent changes the parameter representation by converting distance information into depth map data and using machine learning models that process multiple parameters simultaneously (distance, depth, focus detection results). This parameter transformation enables the system to distinguish between actual subject distance and blocking object distance more effectively, improving focus detection precision while maintaining blocking object identification reliability.
2Quantity of substance
If multiple focus detection regions are used to detect distance information, then more distance data is available for inference, but the influence of erroneous distance information from blocking objects or low-contrast areas increases
Solution Approach 1:
The patent introduces a machine learning model as an intermediary between raw distance information and focus detection results. This intermediary processes multiple distance information pieces, identifies and suppresses erroneous data from blocking objects or low-contrast regions, and produces refined focus detection results. The machine learning model acts as a mediator that filters out harmful information while preserving useful distance data.
Solution Approach 2:
The system implements feedback mechanisms where focus detection results from multiple regions are continuously evaluated and compared. The machine learning model receives feedback from depth map comparisons and adjusts its processing to suppress contributions from regions with blocking objects or detection errors. This feedback loop ensures that erroneous distance information does not compromise overall focus detection precision.
Data Source
AI summary
There is provided an inference apparatus. An inference unit performs inference with use of a machine learning model based on a subject region including a subject within an image obtained through shooting, and on a plurality of distance information pieces detected from a plurality of focus detection regions inside the subject region, thereby generating an inference result indicating a distance information piece corresponding to the subject or a distance information range corresponding to the subject. The machine learning model is a model that has been trained to suppress a contribution made to the inference result by one or more distance information pieces that are not based on the subject among the plurality of distance information pieces.