Image Recognition via 1D Vector Conversion and Subtraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image recognition methods require significant computational resources and processing time, and pose privacy concerns due to the storage of raw images, especially when images are not captured from the same angle as the reference images.
Innovation Solution
An image recognition device and method that converts reference images into 1-dimensional vectors, applies dimension reduction using a random matrix, and calculates coefficients to express input images as a linear sum of reference images, allowing for fast recognition and secure storage by determining the corresponding reference image through subtraction computation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If image recognition is performed by comparing captured images with stored reference images, then authentication can be achieved, but processing time increases and computational resources are consumed
Solution Approach 1:
The patent applies preliminary action by pre-converting reference images into 1-dimensional vectors and storing them in advance. When authentication is needed, the system only needs to compare the converted input image with the pre-converted reference vectors, avoiding the need to process full images during authentication, thus reducing processing time while maintaining accuracy
Solution Approach 2:
The patent extracts essential features from images by converting them into 1-dimensional vectors that capture the most important characteristics. This extraction process removes unnecessary image data while preserving the key information needed for accurate authentication, reducing both processing time and computational requirements
2Measurement precision
If full-resolution reference images are stored for accurate comparison, then recognition precision is improved, but storage requirements and privacy risks increase
Solution Approach 1:
The patent extracts only the essential features from images by converting them into compact 1-dimensional vectors. This extraction maintains the discriminative power needed for precise recognition while dramatically reducing the storage space required, as the vector representation is much more compact than full-resolution images
Solution Approach 2:
The patent changes the representation parameters of images from high-dimensional pixel data to low-dimensional vector data. This parameter transformation preserves the essential information needed for accurate comparison while reducing storage requirements and enhancing privacy protection
3Reliability
If raw reference images are stored in the system, then accurate comparison is enabled, but user privacy protection is compromised
Solution Approach 1:
The patent extracts only the essential mathematical features from images and stores them as 1-dimensional vectors. This extraction removes personally identifiable visual information while preserving the unique characteristics needed for reliable authentication, thus protecting user privacy without compromising authentication reliability
Solution Approach 2:
The patent introduces 1-dimensional vectors as an intermediary representation between the original image and the authentication process. This intermediary form captures the essential identity information while obscuring the actual visual appearance, preventing privacy invasion while maintaining authentication reliability
Data Source
AI summary
An image recognition method and device are disclosed. The disclosed device may include: a reference image storage unit configured to convert reference images into 1-dimensional vectors and store the converted 1-dimensional vectors; an image acquisition unit configured to obtain an input image; an image converter unit configured to convert the input image into a 1-dimensional vector, apply dimension reduction, and calculate coefficients for expressing the input image as a linear sum of vectors forming a reference image for each of the reference images; and a determiner unit configured to determine a reference image corresponding to the input image by performing a subtraction computation of subtracting a value obtained by applying the coefficients calculated for each reference image to a relevant reference image from the converted input image. The invention makes it possible to perform image recognition at a fast speed and lower the likelihood of the users' privacy being invaded.