Featureless Image Categorization Through Unsupervised Low-Pass Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image classification algorithms struggle with categorizing images with amorphous features, such as fire, snow, water, and grassy fields, as they fail to capture these featureless images effectively.
Innovation Solution
The system employs unsupervised learning with multiple low pass filters to progressively blur images, generating multiple versions of both unseen and recorded images. These versions are then hashed and compared to identify matches, allowing for accurate categorization and implementation of responsive actions based on the resiliency status of a device.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional image classification algorithms are used to categorize images with distinct features, then classification accuracy for images with clear boundaries (such as faces, cars, dogs) is improved, but classification fails for images with amorphous features (such as fire, snow, water)
Solution Approach 1:
The patent segments the image classification process into multiple stages: first applying a low-pass filter to create a blurred version for coarse categorization, then using the result to guide selective processing of the original image. This segmentation allows the system to handle both distinct and amorphous features effectively by processing at different levels of detail
Solution Approach 2:
The patent changes the parameter of image blur by applying a low-pass filter to generate a blurred version of the input image. This parameter transformation enables the system to categorize amorphous images by comparing their overall blur patterns rather than relying on sharp feature boundaries
2Measurement precision
If multiple low pass filters are applied to generate multiple versions of images for comparison, then categorization accuracy for amorphous images is improved, but computational complexity and processing time increase
Solution Approach 1:
The patent applies partial action by using only one or a limited number of low-pass filters rather than exhaustive filtering at multiple levels. The system generates a blurred version for categorization purposes and stops when the categorization is achieved, avoiding unnecessary additional filtering steps that would increase computational complexity
3Measurement precision
If multiple low pass filters are applied to generate multiple versions of images for comparison, then categorization accuracy for amorphous images is improved, but processing time increases
Solution Approach 1:
The patent performs preliminary action by first generating a blurred version of the image using a low-pass filter before attempting to categorize. This preliminary blurred version is used to quickly identify potential matches in the database, reducing the need for time-consuming detailed comparisons of the original high-resolution images
Data Source
AI summary
Systems, computer program products, and methods are described herein for image categorization and recognition using unsupervised learning. The present disclosure is configured to receive an unseen image; implement a first low pass filter on the unseen image to generate a first version of the unseen image; implement a second low pass filter on the first version of the unseen image to generate a second version of the unseen image; determine a subset of the second version of the recorded images that matches the second version of the unseen image; retrieve a subset of first version of recorded images associated with the subset of second version of recorded images; and determine a first recorded image from the subset of the first version of the recorded images that matches the first version of the unseen image.