Composite Image Channel Assignment for Motion-Aware Image Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image analysis methods using neural networks struggle with accurately capturing temporal and spatial relationships in moving images and three-dimensional data due to high data capacity, leading to insufficient estimation of fine movements and object attributes.
Innovation Solution
A method involving channel assignment and composite image generation from multiple images with temporal or spatial continuity, followed by inference using a learned model, to enhance the analysis of motion and attributes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If a composite image is generated by simply adding luminance values of a plurality of images, then the data capacity is reduced and processing becomes easier, but the temporal anteroposterior relationship and fine movement information are lost
Solution Approach 1:
The patent divides the composite image into multiple channels, where each channel corresponds to a specific temporal layer. This segmentation allows the neural network to process temporal relationships by analyzing different channels separately while maintaining the benefits of reduced data capacity and simplified processing.
Solution Approach 2:
The patent introduces a temporal dimension by organizing multiple images into separate channels rather than simply stacking them. This dimensional transformation enables the neural network to capture temporal anteroposterior relationships and fine movement information while maintaining processed efficiency.
2Reliability
If moving image data and three-dimensional data are directly input into a neural network, then the analysis can be performed on original data, but convergence of learning and processing capacity are insufficient
Solution Approach 1:
The patent extracts temporal and spatial information from moving image data and three-dimensional data by generating composite images with multiple channels. This extraction transforms the complex data into a format that neural networks can process more efficiently, improving both learning convergence and processing capacity while maintaining analysis accuracy.
Solution Approach 2:
The composite image with multiple channels serves as an intermediary representation between the original moving image data and the neural network. This intermediary format bridges the gap between complex source data and neural network processing capabilities, enabling both high accuracy and efficient processing.
3Quantity of substance
If a composite image is generated from multiple images, then the data capacity is reduced, but the estimation accuracy of action class and fine movements is insufficient
Solution Approach 1:
The patent segments the composite image into multiple channels, each representing a temporal layer. This segmentation preserves fine movement information by allowing the neural network to analyze temporal relationships through channel comparisons, thereby maintaining high estimation accuracy despite reduced data capacity.
Solution Approach 2:
The patent applies local quality by assigning different characteristics to different channels. Each channel has specific temporal information, allowing the neural network to focus on local temporal relationships and fine movements in specific regions, thereby maintaining measurement precision while reducing overall data capacity.
Data Source
AI summary
An example apparatus comprising an image acquisition unit configured to acquire a plurality of images having temporal or spatial continuity; a channel assignment unit configured to assign a channel different from each other to at least a part of gradation information on a color and/or gradation information on brightness that can be acquired from each image of the plurality of images based on a predetermined rule; a composite image generation unit configured to generate one composite image in which gradation information on at least a part of each image can be identified by the channel by extracting and combining gradation information to which the channel is assigned from each of the plurality of images; and an inference unit configured to analyze the composite image and to infer the plurality of images.