Narrow and Wide-Area Image Processing for Lower-Cost Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing learning models incur significant calculation costs and estimation accuracy deteriorates when using high-resolution entire images for training and estimation due to excessive information.
Innovation Solution
The use of narrow-area and wide-area images, where narrow-area images capture detailed features and wide-area images reduce information amount, combined with a learning model trained on these images, to enhance estimation accuracy and reduce calculation costs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If high-resolution entire images are used for training and estimation, then estimation coverage is improved, but calculation costs increase significantly and estimation accuracy deteriorates
Solution Approach 1:
The patent divides the high-resolution entire image into multiple low-resolution partial images (first through fourth images) and processes them separately through the learning model. This segmentation approach reduces the computational load for each individual image processing operation while maintaining comprehensive coverage of the entire target through aggregation of results from all partial images.
2Measurement precision
If high-resolution entire images are used for training and estimation, then estimation coverage is improved, but estimation accuracy deteriorates due to excessive information
Solution Approach 1:
The patent segments the high-resolution entire image into multiple low-resolution partial images, which reduces information overload in each individual processing operation. The learning model processes these simplified partial images separately, avoiding the computational and accuracy issues associated with processing the complete high-resolution image as a single unit.
Solution Approach 2:
The patent applies partial action by processing only the necessary partial images through the learning model rather than the entire high-resolution image. This selective processing approach maintains sufficient information for accurate estimation while avoiding the detrimental effects of processing excessive information in a single operation.
3Measurement precision
If the entire high-resolution image is processed through the learning model, then comprehensive information is captured, but processing time increases significantly
Solution Approach 1:
The patent segments the image processing task into multiple parallel operations on partial images. By dividing the high-resolution entire image into separate low-resolution partial images that can be processed simultaneously through the learning model, the system significantly reduces total processing time while maintaining comprehensive information capture through aggregation of partial results.
Data Source
AI summary
An information processing apparatus includes an image acquiring unit that acquires a plurality of narrow-area images which are respectively images of parts of a target image and include mutually different ranges of the target image and a wide-area image which is an image having an information amount of the target image and includes a range of the target image larger than that included in each of the plurality of narrow-area images, and an information acquiring unit that acquires information correlated with the target image from a learning model to which the plurality of narrow-area images and the wide-area image are input.