Trace Transform Image Identification Using Mask Function
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current image identification techniques face challenges in achieving a false alarm rate significantly lower than 1 ppm and maintaining detection rates above 98%, especially in the presence of image modifications such as histogram equalization and cropping.
Innovation Solution
The method involves extracting visual identification features from the Trace transform of an image, spatially restricting the Trace transform to derive additional binary component identifiers, which are combined with existing descriptors to enhance robustness and performance, and using a mask function to compute the circus function from bands in the Trace domain, thereby improving detection rates to 99.80% at a reduced false alarm rate of 0.1 ppm.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional image identification methods are used, then computational complexity is reduced and descriptor size is reduced, but false alarm rate cannot be lowered below 1 ppm level
Solution Approach 1:
The patent segments the Trace transform representation into multiple subsets of lines, extracting binary component identifiers from each subset. This segmentation allows the system to process and compare multiple restricted representations rather than one complete representation, improving reliability by providing multiple opportunities for correct identification while maintaining computational efficiency through localized processing.
Solution Approach 2:
The patent extracts binary component identifiers from subsets of lines rather than processing the entire Trace transform. This partial action approach reduces computational complexity per component while generating multiple identifiers that can be combined, ultimately achieving lower false alarm rates through the cumulative effect of multiple partial extractions.
2Reliability
If traditional image identification methods are used, then detection rate is maintained above 98%, but robustness to histogram equalisation and image cropping is insufficient
Solution Approach 1:
The patent applies spatial restriction to create subsets of lines from the Trace transform, where each subset focuses on specific regions or orientations. This local quality approach ensures that modifications in one part of the image (such as cropping or histogram equalisation) do not completely destroy the identifier, as other restricted representations remain intact and can still provide reliable detection.
Solution Approach 2:
The patent combines multiple binary component identifiers extracted from different subsets of lines into a composite identifier. This composite approach enhances robustness because the combined identifier leverages information from multiple independent subsets, making it more resistant to various image modifications than any single subset alone.
3Reliability
If additional binary component identifiers are extracted from spatially restricted Trace transform, then detection rate increases to 99.80%, but false alarm rate reduction to 0.1 ppm requires more processing
Solution Approach 1:
The patent segments the Trace transform into multiple subsets and extracts binary component identifiers from each segment in parallel. This segmentation enables efficient processing by distributing the computational workload across multiple independent extractions, achieving high detection rates through the combination of multiple identifiers without proportionally increasing overall processing time.
Solution Approach 2:
The patent performs spatial restriction and subset creation as preliminary actions before extracting binary component identifiers. By pre-organizing the Trace transform into meaningful subsets based on spatial or directional criteria, the system prepares the data structure to enable rapid extraction of multiple identifiers, improving processing efficiency while maintaining high detection rates.
Data Source
Figure 1(a)~1(f)
Figure 2
Figure 3
AI summary
A method and apparatus for deriving a representation of an image is described. The method involves processing signals corresponding to the image. A two-dimensional function of the image, such as a Trace transform (T (d, θ)), of the image using at least one functional T, is derived and processed using a mask function (β) to derive an intermediate representation of the image, corresponding to a one-dimensional function. In one embodiment, the mask function defines pairs of image bands of the Trace transform in the Trace domain. The representation of the image may be derived by applying existing techniques to the derived one-dimensional function.